University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2022 Rachel K. Straub

DEVELOPMENT OF A MOVEMENT PERFORMANCE ASSESSMENT TO
PREDICT ACL RE-INJURY

by

Rachel K. Straub

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)

August 2022

ii
DEDICATION

This dissertation is dedicated to the millions of people worldwide who are
severely disabled from Myalgic Encephalomyelitis/Chronic Fatigue Syndrome
(ME/CFS), which is a complex multi-system disease characterized by debilitating
fatigue, cognitive dysfunction, orthostatic intolerance, pain, and post-exertional
malaise (i.e., worsening of symptoms after minimal exertion, which can be as simple
as taking a shower). Unfortunately, this disease has no cure nor FDA approved
treatment. Despite my struggles with this illness for over 9 years, I have been immensely
blessed with what I have been able to accomplish. It is only because of the unrelenting
support of numerous individuals, most importantly my parents and academic advisor,
that this dissertation has been possible. Although the contents of this dissertation have
no bearing on ME/CFS, I remain hopeful that this dissertation represents a
steppingstone in my life-long mission to improve healthcare and treatment outcomes
for a variety of medical conditions, including ME/CFS.

iii
ACKNOWLEDGEMENTS

I would like to begin by expressing my sincerest apology to my advisor and the
entire department at USC in the Division of Biokinesiology and Physical Therapy.
When I entered the PhD program in the Fall of 2017, it was under the assumption I
was finally healthy after a 4.5-year battle with severe ME/CFS. However, my body
started to break down my first semester, and my attempt to maintain a mirage that
everything was fine was not sustainable. Therefore, I am forever indebted to my advisor
and the department, including Dr. Gordon, for supporting my ongoing recovery
throughout my entire duration as a PhD student. In addition, I am also indebted for
the financial assistance I received from the Division of Biokinesiology and Physical
Therapy and the Honor Society of Phi Kappa Phi, which helped make this work
possible.
Although my journey as a PhD student has been overshadowed by my health
struggles, being part of this program (and the research I have been doing) has been
something that I have loved. It’s been a very humbling experience being in a PhD
program while struggling with basic function, which was never part of the plan. Despite
all this, being part of this program has in many ways helped me heal. There is nothing
I regret (or would change) if I had the ability to turn back time. I have had the privilege
of being part of a research project from the ground up, building a research lab from
scratch, and becoming the first student in the division to use the Simi Motion system
for 3D motion analysis. In addition, I have been a part of some amazing projects, have
met some amazing people, and have developed skills that will hopefully, one day, allow
me to make the world a better place. It is only because of the support and assistance
from numerous people, who are detailed below, that I have been able to complete my
PhD.
There are not enough words in the English language to express my gratitude to
my academic advisor, Dr. Christopher Powers. Although our time together started with

iv
me as a master’s student, my health caused me to leave the program prematurely and
finish my degree from home. The plan was always for me to return to USC for a PhD
once I was healthy, but neither of us expected this would take 4.5 years. Dr. Powers
patiently waited for my recovery and put my potential project on hold for years on end.
And then, once I returned, although my health started to deteriorate, he continued to
work with me. Dr. Powers showered me with helpers and gifted me with a dissertation
project where 75% of it could be completed from home. Without this unrelenting
support, none of this project would have been possible. It is only because of the
generosity, patience, and kindness of one person, Dr. Christopher Powers, that my
dream of obtaining a PhD in Biokineisology has come to fruition.
I would like to sincerely thank my committee members for their continuous
support throughout my entire duration as a PhD student. Dr. Keim, thank you for
always being available remotely for literally anything related to statistics and for coming
out of retirement to serve on my committee. Even though we have never met in person,
having you serve on my committee remotely has meant more to me than I can express
with words. Dr. Salem, thank you for always challenging me to understand the “why”
behind my methodology and for encouraging me to understand the basic elements of
my dissertation that I may have overlooked. Your continual encouragement and
support, despite my ongoing health challenges, are appreciated more than you will ever
know. It’s been both a privilege and an honor to have you on my committee. Your
kind spirit has aided my recovery, and for this this, I am forever grateful. Dr. Sigward,
thank you for the endless hours of in-person tutoring on biomechanics 101 that forced
me to go back to the drawing board to understand the past literature that formed the
foundation of my dissertation. Your basic teachings helped improved the overall
content of my dissertation, so I am forever grateful. And lastly, I extend my deepest
and sincerest thanks to Dr. Michener, who has been a continual force behind all
elements within my dissertation. Your ongoing dedication, from my proposal to the
contents of my dissertation to enhancing my knowledge basis, has been above and

v
beyond. In addition, your emotional support and encouragement, and your willingness
to help me in any way possible (including with events following graduation), has played
a pivotal role in my recovery, which has meant more to me than I can ever express.
All my committee members have generously shared their time and expertise to help
make my dissertation successful, and for this, I am forever grateful.
I could not have completed this dissertation without the help of endless
volunteers. I would like to express my deepest gratitude to Alex Horgan, who acted as
my personal assistant for my first 1.5 years as a PhD student. She served as a pilot
subject, uploaded files for me at the Movement Performance Institute (MPI), and
scored over 100 patients for my dissertation. And once she returned home to Ireland,
she acted as a recruiter, sending me over 10 patients. I would also like to thank Adam
Barrack, who uploaded files for me at MPI, served as a pilot subject, and did the
physical labor required to get the MPI lab up and running (from making the marker
clusters to transporting the force plate). He also trained my first dissertation subject on
the MPI protocol and consulted with Simi Motion on technical issues (and sent me
the updates), so I didn’t have to make any additional trips to the facility. And finally,
my deepest thanks to all the volunteers who assisted me with data collection: Chris
Miller, Leana Mosesian, Justin Kouyoumdjian, Franklin Zhuang, Ruben Cuellar,
Aaron Tadle, Tim Brown, and Chenyong Xiang. Additional thanks go out to the
parents and coaches who sent me more patients than I was capable of testing: Ronan
Wall, Michelle Yarger, Fonda Kim-Tokushige, and Maxine Lewis.
Finally, none of this would have been possible without the ongoing love and
support of my parents. Their unwavering support, even in my darkest hours, is the only
reason I am alive today. They have literally nursed me back to health for over 9 years,
sometimes even spoon feeding me when I haven’t been capable of feeding myself, and
for this I am eternally grateful.

vi
TABLE OF CONTENTS

DEDICATION ____________________________________________________ ii
ACKNOWLEDGEMENTS ___________________________________________ iii
LIST OF TABLES _________________________________________________ viii
LIST OF FIGURES ________________________________________________ ix
ABSTRACT _______________________________________________________ xii
CHAPTER 1: OVERVIEW ___________________________________________ 1
CHAPTER 2: BACKGROUND AND SIGNIFICANCE _____________________ 3
Statement of the Problem __________________________________________ 3
Movement Impairments and ACL Injury Risk __________________________ 4
Movement Impairments Associated with Initial ACL Injury Risk __________ 4
Movement Impairments Associated with Secondary ACL Injury Risk _______ 5
Summary _____________________________________________________ 8
Currently Available Movement Assessments ____________________________ 8
Introducing the Movement Performance Assessment (MPA) _______________ 9
MPA Movement Domains _______________________________________ 10
Summary ______________________________________________________ 12
CHAPTER 3: CLINICAL ESTIMATION OF THE USE OF THE HIP AND
KNEE EXTENSORS DURING ATHLETIC MOVEMENTS USING 2D VIDEO
_________________________________________________________________ 14
Introduction ___________________________________________________ 14
Methods ______________________________________________________ 15
Participants __________________________________________________ 15
Sample-Size Calculation ________________________________________ 16
Instrumentation _______________________________________________ 16
Procedures ___________________________________________________ 17
Data Analysis _________________________________________________ 18
Statistical Analysis _____________________________________________ 19
Results ________________________________________________________ 20
Discussion _____________________________________________________ 21
Conclusion ____________________________________________________ 23
CHAPTER 4: ESTIMATION OF VERTICAL GROUND REACTION FORCE
PARAMETERS DURING ATHLETIC TASKS USING 2D VIDEO __________ 24
Introduction ___________________________________________________ 24
Methods ______________________________________________________ 26
Participants __________________________________________________ 26
Sample-Size Calculation ________________________________________ 26
Instrumentation _______________________________________________ 26
Procedures ___________________________________________________ 27
Data Analysis _________________________________________________ 27
Statistical Analysis _____________________________________________ 28
Results ________________________________________________________ 29
Discussion _____________________________________________________ 32
Conclusion ____________________________________________________ 35
CHAPTER 5: UTILITY OF 2D VIDEO ANALYSIS FOR ASSESSING FRONTAL
PLANE TRUNK AND PELVIS MOTION DURING STEPPING, LANDING,
AND CHANGE IN DIRECTION TASKS: A VALIDITY STUDY ____________ 36
Introduction ___________________________________________________ 36

vii
Methods ______________________________________________________ 37
Participants __________________________________________________ 37
Instrumentation _______________________________________________ 38
Procedures ___________________________________________________ 38
Data Analysis _________________________________________________ 39
Statistical Analysis _____________________________________________ 40
Results ________________________________________________________ 41
Descriptive Data ______________________________________________ 41
Correlation and Agreement between 2D and 3D Frontal Plane Pelvis Angles 41
Correlation and Agreement between 2D and 3D Frontal Plane Trunk Angles 43
Discussion _____________________________________________________ 45
Conclusion ____________________________________________________ 48
CHAPTER 6: DOES THE 2D FRONTAL PLANE PROJECTION ANGLE
PREDICT FRONTAL PLANE KNEE MOMENTS DURING STEPPING,
LANDING, AND CHANGE OF DIRECTION TASKS? ___________________ 49
Introduction ___________________________________________________ 49
Methods ______________________________________________________ 51
Participants __________________________________________________ 51
Sample-Size Calculation ________________________________________ 51
Instrumentation _______________________________________________ 51
Procedures ___________________________________________________ 52
Data Analysis _________________________________________________ 52
Statistical Analysis _____________________________________________ 53
Results ________________________________________________________ 54
Relationship between FPPA and Peak Frontal Plane Knee Moment ______ 55
Relationship between FPPA and Average Frontal Plane Knee Moment ____ 56
Relationship between FPPA and Frontal Plane Knee Moment at Peak Knee
Flexion ______________________________________________________ 57
Discussion _____________________________________________________ 58
Conclusion ____________________________________________________ 61
CHAPTER 7: PREDICTION OF ACL RE-INJURY IN FEMALES USING 2D
VIDEO: A RETROSPECTIVE CASE-CONTROL STUDY __________________ 62
Introduction ___________________________________________________ 62
Methods ______________________________________________________ 63
Participants __________________________________________________ 63
Sample-Size Calculation ________________________________________ 64
Procedures ___________________________________________________ 64
Data Analysis _________________________________________________ 67
Statistical Analysis _____________________________________________ 68
Results ________________________________________________________ 70
Clustering ____________________________________________________ 70
Movement Domain Prediction ____________________________________ 75
Discussion _____________________________________________________ 78
Conclusion ____________________________________________________ 81
CHAPTER 8: SUMMARY AND CONCLUSIONS ________________________ 83
Clinical Implications _____________________________________________ 86
Limitations and Direction for Future Research _________________________ 88
Conclusion ____________________________________________________ 90
BIBLIOGRAPHY: __________________________________________________ 92
APPENDIX: SAMPLE SURVEY _____________________________________ 107

viii
LIST OF TABLES

Table 2-1. Movement Risk Factors for Initial ACL Injury ............................................ 5

Table 2-2. Movement Risk Factors for Secondary ACL Injury ..................................... 7

Table 2-3. Overview of Movement Performance Assessment (MPA) .......................... 10

Table 3-1. Characteristics of Study Participants, Mean (SD) ..................................... 16

Table 3-2. Description of Athletic Tasks ..................................................................... 18

Table 3-3. Average Hip/Knee Extensor Moment and 2D Trunk-Tibia Angles at Peak
Knee Flexion .............................................................................................................. 20

Table 4-1. Vertical Ground Reaction Force Variables and 2D Thigh Angle, Mean
(SD) ........................................................................................................................... 29

Table 5-1. Frontal Plane Angles for Pelvis and Trunk using 2D Video and 3D Motion
Analysis, Mean (SD) .................................................................................................. 41

Table 7-1. Matching Criteria for Cases & Controls* ................................................... 65

Table 7-2. Patient Demographics (N = 84)* ............................................................... 67

Table 7-3. 2D Measurements (degrees) within Clusters for Each Movement Domain
(N = 84) .................................................................................................................... 74

Table 7-4. Separate Logistic Regression Models for the 5 MPA Movement Domains
to Predict ACL Re-injury in Female Athletes ............................................................ 75

Table 7-5. Statistics for Model Improvement with Addition of Knee Strategy Domain
to Baseline Model* .................................................................................................... 77

ix
LIST OF FIGURES

Figure 2-1. Knee strategy is scored using the trunk-tibia angle (+trunk more forward
than tibia, -tibia more forward than trunk). Scored during all 6 tasks. ...................... 10

Figure 2-2. Shock absorption is scored using the thigh angle (+increased hip and/or
knee flexion). Scored during 5 of 6 tasks (all except step down). .............................. 11

Figure 2-3. Pelvis stability is scored using pelvis tilt (+drop, -rise). Scored during 5 of
6 tasks (all except drop jump). ................................................................................... 11

Figure 2-4. Trunk stability is scored using trunk lean (+ipsilateral lean, -contralateral
lean). Scored during 5 of 6 tasks (all except drop jump). .......................................... 12

Figure 2-5. Knee stability is scored using the frontal plane projection angle (+valgus, -
varus). Scored during all 6 tasks. ................................................................................ 12

Figure 3-1. Linear regression models to predict the average hip/knee extensor moment
for each task based on the 2D trunk-tibia co-varied for body mass. Predicted 3D
average HKR values were calculated by regressing the 2D trunk-tibia and body mass
on the observed average HKR. The beta coefficient for the 2D trunk-tibia for all
models was positive. Abbreviations. HKR: hip/knee extensor moment ratio. ............ 21

Figure 4-1. Time-normalized vGRF data for the 5 tasks evaluated. Error bars
represent 1 SD. Abbreviations. vGRF: vertical ground reaction force. ...................... 30

Figure 4-2. Linear regression models for each task separately to predict vGRF 1st
peak. vGRF normalized to body mass and approach velocity for all tasks except drop
jump, which was normalized to body mass only. Abbreviations. vGRF: vertical
ground reaction force. ................................................................................................ 31

Figure 4-3. Linear regression models for each task separately to predict vGRF impulse
to 1st peak. vGRF normalized to body mass and approach velocity for all tasks
except drop jump, which was normalized to body mass only. Abbreviations. vGRF:
vertical ground reaction force. .................................................................................... 32

Figure 5-1. Correlation models for the 2D and 3D frontal plane pelvis angles for each
task. ............................................................................................................................ 42

Figure 5-2. Bland Altman plots comparing 2D vs. 3D frontal plane pelvis angles for
each task. Upper and lower dotted lines represent 95% limits of agreement. Solid line
represents bias or mean difference. Positive mean values indicate pelvis drop; negative
mean values indicate pelvis rise. Abbreviations. MD: mean difference. ..................... 43

Figure 5-3. Correlation models for the 2D and 3D frontal plane trunk angles for each
task. ............................................................................................................................ 44

Figure 5-4. Bland Altman plots comparing 2D vs. 3D frontal plane trunk angles for
each task. Upper and lower dotted lines represent 95% limits of agreement. Solid line
represents bias or mean difference. Positive mean values indicate ipsilateral lean;
negative mean values indicate contralateral lean. Abbreviations. MD: mean
difference. .................................................................................................................. 45

x

Figure 6-1. Average FPPA and moment variables for the 6 tasks evaluated. Error bars
represent 1 SD. .......................................................................................................... 54

Figure 6-2. Time-normalized frontal plane knee moment data for the 6 tasks
evaluated. Error bars represent 1 SD. Positive values represent knee valgus moments.
................................................................................................................................... 55

Figure 6-3. Linear regression models to predict the peak frontal plane knee moment
for each task. .............................................................................................................. 56

Figure 6-4. Linear regression models to predict the average frontal plane knee
moment for each task. ................................................................................................ 57

Figure 6-5. Linear regression models to predict the frontal plane knee moment at
maximum knee flexion for each task. ......................................................................... 58

Figure 7-1. Flow chart of 345 female athletes contacted to complete survey, which
resulted in 23 cases and 61 controls. RTS: return to sport. ........................................ 66

Figure 7-2. The MPA is comprised of 5 separate movement domains, each which
contains 2D metrics for 5-6 tasks. In total, 33 angular measures are required, which
results in 27 measures for data analysis. ..................................................................... 68

Figure 7-3. Optimal number of clusters for each movement domain was determined
by finding the cluster solution with the maximum silhouette value. The silhouette
value was maximized with a 2-cluster solution (k=2) for all movement domains. ..... 71

Figure 7-4. Visualization of k-means clustering output (k=2) after using principial
component analysis to reduce multi-dimensional data to two-dimensions. K-means
clustering was performed separately for each MPA movement domain, with 84
athletes. ...................................................................................................................... 72

Figure 7-5. Cluster characteristics in female athletes post ACLR. K-means clustering
(k=2) was performed separately for each MPA movement domain with 5-6 variables
(tasks). Each row (color) represents a movement domain, and each column
represents a task. Athletes are consistent across rows only. C1: Cluster 1; C2: Cluster
2. The characterization for each cluster (which differs based on movement domain) is
shown in Table 7-3. Knee strategy is scored using the trunk-tibia angle (+trunk more
forward than tibia, -tibia more forward than trunk). Shock absorption is scored using
the thigh angle (+increased hip and/or knee flexion). Knee stability is scored using
the frontal plane projection angle (+valgus, -varus). Trunk stability is scored using
trunk lean (+ipsil, -contra). Pelvis stability is scored using pelvis tilt (+drop, -ipsil).
................................................................................................................................... 73

Figure 7-6. Within the knee strategy domain, the high knee extensor bias subgroup
(defined by a lower 2D trunk-tibia angle across tasks) was at increased risk for ACL
re-injury compared to the low knee extensor bias subgroup. ..................................... 76

Figure 7-7. Risk of probability for ACL re-injury in cases increased 4% on average
when the knee strategy domain was added to the baseline model with 3 adjustment
factors. The change in risk ranged from -17% to 12%. The shift in ACL re-injury risk

xi
was always positive (5% to 12%) when in the presence of a high knee extensor bias
strategy (n=17). .......................................................................................................... 77

xii
ABSTRACT

Approximately 1 in 4 young athletes who return to a high-risk sport after primary
ACL reconstruction (ACLR) will go on to sustain another ACL injury.
109,156

Interestingly, 60-70% of ACL injuries (initial or secondary) occur as a result of non-
contact mechanisms.
2,51
The fact that movement-related impairments are thought to
underlie non-contact ACL injuries,
59,70,71,77,79,80,110,119
highlights the need to assess
movement behavior to better determine readiness to return to sport.
The lack of a comprehensive clinical assessment to quantify an athlete’s
readiness to return to sport following ACLR led to the development of the video-based
Movement Performance Assessment (MPA). Although the MPA has been developed
and is being used clinically, it remains unknown if the 2D measures of the MPA
represent/predict 3D measures. In addition, it is not known whether the movement
constructs that comprise the MPA are important for predicting ACL re-injury. This
dissertation sought to answer these questions as a first step in establishing the MPA as
a potential clinical tool to assess an athlete’s readiness to return to sport following
ACLR.
The purpose of Chapters 3-6 (Aim 1) was to determine if the 2D movement
variables established for the MPA are representative of the 3D variables related to ACL
injury risk based on laboratory-based studies. Thirty-nine healthy athletes (15 males
and 24 females) performed 6 tasks conceptualized for the MPA (step down, drop jump,
lateral shuffle, deceleration, triple hop, and side-step-cut) while 3D kinematics/kinetics
and 2D video were collected simultaneously. Specific 2D angular measurements and
the corresponding 3D kinematic and kinetic variables were quantified during the
deceleration/lowering phase of each task. Linear regression or correlation analysis was
used to assess the relationship between 2D and 3D metrics. Agreement between 2D
and 3D angles was assessed using Bland Altman plots, when relevant.

xiii
Validation of the kinetic constructs of the MPA (knee strategy and shock
absorption) are presented in Chapters 3 and 4. The purpose of Chapter 3 was to
determine whether the difference between sagittal plane trunk and tibia orientations
obtained from 2D video (2D trunk-tibia) at peak knee flexion could be used to predict
the average hip/knee extensor moment ratio during the deceleration/lowering phase of
each of the 6 MPA tasks. For each task, an increase in the 2D trunk-tibia angle
(representing a more forward trunk relative to the tibia) predicted an increase in the
average hip/knee extensor moment ratio when adjusted for body mass (all p < .013, R
2

= 0.17-0.77). The purpose of Chapter 4 was to determine whether the 2D thigh angle
at peak knee flexion relative to the vertical could be used to predict the peak vertical
ground reaction force (vGRF) and vGRF impulse during the tasks that involved impact
with the ground (all except step down). For all impact tasks except for cutting, an
increase in the 2D thigh angle (which is representative of increased hip and knee
flexion) predicted a lower peak vGRF (R
2
= 0.17 to 0.47, all p < 0.01). However, an
increase in the 2D thigh angle predicted a lower vGRF impulse for all MPA impact
that involved impact (R
2
= 0.13 to 0.39, all p < 0.025).
Validation MPA constructs that involved kinematic constructs (trunk stability,
pelvis stability, and knee stability) are presented in Chapters 5-6. The purpose of
Chapter 5 was to determine whether the 2D frontal plane angles of the trunk and pelvis
at peak knee flexion were associated with the corresponding 3D angles during the MPA
tasks that involved single limb contact with the ground. In addition, agreement between
2D/3D angles was assessed. 2D and 3D frontal plane angles for all tasks were correlated
in a positive direction at the pelvis (r = 0.54 to 0.73, all p < 0.001) and trunk (r = 0.81
to 0.92, all p < 0.001). Absolute agreement in the frontal plane for all tasks and angles
was below 5°. However, the 95% limits of agreement across tasks ranged from -12.8°
to 21.3° for the pelvis and -11.8° to 9.4° for the trunk.
The purpose of Chapter 6 was to establish whether the 2D frontal plane
projection angle (FPPA) at peak knee flexion could be used to predict frontal plane

xiv
knee moments (peak moment, average moment, moment at peak knee flexion) during
each of the 6 MPA tasks. An increased FPPA (inward collapse of the knee) significantly
predicted the peak frontal plane knee moment for 2 tasks (deceleration and side-step-
cut, R
2
= 0.12 to 0.25), average frontal plane knee moment for 5 tasks (drop jump,
shuffle, deceleration, triple hop, side-step-cut, R
2
= 0.15 to 0.40), and frontal plane
knee moment at peak knee flexion for 5 tasks (drop jump, shuffle, deceleration, triple
hop, side-step-cut, R
2
= 0.16 to 0.45).
Following concurrent validation of the 2D MPA metrics, Chapter 7 sought to
determine if the movement domains evaluated as part of the MPA were relevant for
predicting non-contact ACL re-injury (ipsilateral or contralateral). Female athletes who
previously underwent ACL reconstruction (ACLR) (N=345) and who had previously
undergone return to sport testing using the MPA were surveyed. The survey response
rate was 53%. Females who sustained an ipsilateral or contralateral ACL re-injury
(non-contact) within 36 months of returning to sport were considered as cases (n=23)
and matched with 2-3 non-injured controls (n=61) based on age, graft type, sport level,
and athletic exposures.
Cluster analysis was conducted to separate female athletes in each movement
domain into 2 subgroups, which were operationally defined as “high injury risk” and
“low injury risk.” The underlying 2D angular measurements between subgroups differed
significantly from each other (all p < 0.034), with a consistent pattern across tasks.
Results from logistic regression analysis revealed that only the knee strategy movement
domain was predictive of ACL re-injury. Compared to the “low knee extensor bias”
subgroup (defined by high 2D trunk-tibia angles across tasks), the odds of ACL re-
injury were increased in the “high knee extensor bias” subgroup (defined by low 2D
trunk-tibia angles across tasks) (adjusted OR = 3.19, 95% CI: 1.02, 9.96, p = 0.046).
A receiver operating characteristic curve showed an area under the curve of 78%,
indicating fair prediction accuracy.

xv
Taken together, the results of this dissertation support the use of 2D video
analysis across a wide range of athletic tasks to quantify movement impairments that
have been hypothesized to be associated with elevated risk of ACL re-injury.
Specifically, 2D measures to quantify pelvis stability, trunk stability, knee stability,
shock absorption, and knee strategy can be used as reasonable clinical surrogates for
more complex 3D measures obtained in a laboratory setting. In terms of predicting
ACL re-injury (ipsilateral or contralateral), the movement domain that quantified use
of the knee extensors relative to the hip extensors (knee strategy) was found to be
relevant, with female athletes with a high knee extensor bias being at increased risk
compared to those with a low knee extensor bias. Future studies are needed to verify
these injury prediction results in prospective studies, establish task importance for
quantifying knee strategy, and determine cutoffs that can be used in the clinical setting
to distinguish between a high vs. low knee extensor bias.

1
CHAPTER 1:
OVERVIEW

Approximately 1 in 4 young athletes who return to a high-risk sport after primary
ACL reconstruction (ACLR) will go on to sustain another ACL injury.
109,156
Young
athletes who return to sport have a 30 to 40 times greater risk of sustaining a second
ACL injury compared to their uninjured counterparts.
156
Prospective studies have
identified specific movement impairments at the trunk, pelvis, hip, and knee that are
predictive of initial ACL injury.
59,77,79,80
More importantly, movement impairments at
the trunk, pelvis, hip, and knee can predict a second ACL injury.
70,71,110,119
There is a
need for a clinically feasible assessment that identifies movement impairments related
to primary and secondary ACL injury risk.
One clinic-based movement assessment used to assess ACL injury risk is the
Landing Error Scoring System (LESS).
105
However, the ability of LESS scores to
predict initial ACL injury is limited to youth soccer, with no predictive capacity in high
school and collegiate athletes.
104,140
A shortcoming of LESS is the fact that only a double
limb landing task is evaluated. This limitation is notable as ACL injuries typically occur
during single-limb, multidirectional tasks.
4,29
The lack of a comprehensive clinical
assessment of patients who have experienced an ACL injury led to the development of
the video-based Movement Performance Assessment (MPA). Unlike the LESS that
evaluates a single task, the MPA examines movement quality during range of sport-
specific tasks typically associated with ACL injury. Movement quality is quantified in
terms of 5 movement domains that have been hypothesized to contribute to ACL injury
(knee strategy, shock absorption, knee stability, pelvis stability, and trunk stability).
The primary objective of this dissertation was to establish the utility of the MPA
as a return to sport assessment for persons who have undergone ACLR. To achieve this
objective, this dissertation first examined whether the 2D variables developed for the
MPA are representative of three-dimensional (3D) laboratory-based measures of “at

2
risk” movement behavior for ACL injury. Second, this dissertation sought to establish
if the movement domains that are evaluated as part the MPA are relevant for predicting
ACL re-injury. Data collected as part of this dissertation is the first step in the
development of a valid clinic-friendly assessment to aid in the identification of persons
at risk for re-injury following ACLR.

SPECIFIC AIM 1 (Concurrent Validation): Determine if the 2D movement variables
established for the MPA are representative of the 3D variables related to ACL injury
risk based on prior literature. To achieve this aim, 3D kinematics and kinetics, along
with 2D high speed video, were collected simultaneously during the movement tasks
that have been conceptualized for the MPA.
Hypothesis 1: The 2D kinematic metrics established for each of the movement
domains within the MPA will be able to predict the corresponding 3D kinematic and/or
kinetic variables across all movement tasks.

SPECIFIC AIM 2 (Injury Prediction Modeling): Determine if the movement domains
evaluated as part of the MPA are relevant for predicting non-contact ACL re-injury
(ipsilateral or contralateral). To achieve this aim, a database of 2D videos obtained
from female athletes who underwent return to sport testing using the MPA following
ACLR were analyzed. Re-injury status was determined using an online survey.
Hypothesis 2a: For each of the movement domains that comprise the MPA, a
subgroup (or cluster) of female athletes who exhibit a consistent high-risk movement
profile across tasks can be identified.
Hypothesis 2b: For each of the movement domains, athletes assigned to the high-
risk subgroup will have a greater risk of ACL re-injury compared to those assigned to
lower risk subgroups.

3
CHAPTER 2:
BACKGROUND AND SIGNIFICANCE

Statement of the Problem
Although sport participation is linked to a multitude of health benefits,
41,42
knee
injuries are common among those who participate in athletics and can negatively
impact sport performance. Ligament damage is the most common knee pathology, with
the ACL being one of the most frequently injured ligamentous structures.
23
Sports that
require frequent jumping, landing, and cutting (e.g., soccer, volleyball, and basketball)
are among the sports with the highest ACL injury rates.
2,49,65
In addition, females have
higher rates of initial ACL injury than males after controlling for sport.
2,49,65

Unfortunately, sustaining an initial ACL injury results in a 30 to 40 times greater
risk of sustaining a second ACL injury compared to their uninjured counterparts.
156
Both initial and secondary ACL injuries are similar in that 60-70% occur as a result of
non-contact mechanisms.
2,51
Although females are at increased risk of initial ACL injury
compared with males when playing the same sport, sex differences in risk for
experiencing a second ACL injury are negligible.
107

Laboratory studies have identified specific movement impairments at the trunk,
pelvis, hip, and knee that are predictive of initial ACL injury.
59,77,79,80
More importantly,
a second ACL can be predicted by movement impairments at the trunk, pelvis, hip,
and knee.
70,71,110,119
There is a need for a clinically feasible movement assessment to (1)
identify movement impairments to better define readiness to return to sport, and to (2)
identify those movement deficits related to risk of a subsequent ACL injury following
ACL reconstruction. This is important as current return to sport criteria following
ACLR focuses primarily on measures of knee strength and physical performance (i.e.,
hopping distance, etc.).
6
The following section will review the current state of research
regarding movement impairments and ACL injury risk.

4
Movement Impairments and ACL Injury Risk
Movement Impairments Associated with Initial ACL Injury Risk
Movement impairments at the hip, knee, and pelvis have been reported to be
associated with increased incidence of primary ACL injury, all in female
populations.
59,77,79,80
The majority of studies that have identified movement impairments
as risk factors for primary ACL employed 3D motion analysis (primarily during a drop
jump) (Table 2-1).
59,77,79,80
More specifically, 3D movement risk factors based on
prospective studies during a drop jump include increased knee valgus angles (initial
contact and peak),
59
increased peak knee valgus moments,
59
increased peak knee
extensor moments,
77
decreased hip flexion range of motion (initial contact to peak),
77

and decreased peak knee flexion.
79
During a standing knee lift, 3D pelvis tilt predicts
ACL injury.
80

It is worth noting that studies are conflicting regarding movement risk factors
for initial ACL injury. While Hewett et al. reported knee valgus to be a predictor of
primary ACL injury,
59
more recent work by Leppanen et al.
78
was unable to confirm
this finding. In addition a recent systematic review concluded that knee valgus
angles/moments do not appear predictive of ACL injury.
26
However, the evidence in
support of this conclusion predominantly was based on studies utilizing a double-leg
drop jump, a task in which ACL injuries do not readily occur.
4,29
The fact that athletes
change activity 500-3000 times over the course of a competition,
147
highlights the
limited utility of movement assessments that do not evaluate tasks that involve a change
in direction. The ability of knee valgus to predict ACL injury would appear to require
assessments that include single-limb multidirectional tasks during which most ACL
injuries occur.
4,29

In the study of Hewett et al.,
59
sagittal plane knee flexion was not found to be a
predictor of initial ACL injury, which conflicts with the more recent results of
Leppanen et al.
78
Based on the most robust 3D movement prospective study to date
(710 female athletes screened), movement impairments during the drop jump (frontal

5
and sagittal planes) did not predict initial ACL injury.
75
Differences in methodology
among studies likely account for the disparities among studies. For example, Krosshaug
et al.
75
followed 710 athletes over the course of 7 years, and even though they adjusted
for exposure, the extended follow-up calls into question the validity of these findings
as movement changes with time. In addition, Krosshaug et al. excluded athletes with
previous ACL injuries from their analysis,
75
while Hewett et al.
59
and Leppanen et al.
79

did not.

Movement Impairments Associated with Secondary ACL Injury Risk
Four studies have evaluated biomechanical risk factors during various athletic
tasks as predictors of secondary ACL injury (ipsilateral, contralateral, or
combination).
70,71,110,119
Similar to initial injury, movement impairments at the trunk,
pelvis, hip, and knee have been shown to predict a second ACL injury, but a wider
variety of tasks have been examined predominantly with 3D motion analysis (Table 2-
2).
Table 2-1. Movement Risk Factors for Initial ACL Injury

Author N Injured (%) Task Endpoint Follow-up
Time
Significant Risk Factors*
Hewett
59
205
(female)
9 (4.4%)** Drop
Jump

NC ACL
tear
13 months
• á peak knee valgus
moment
• á knee valgus (IC
& peak)

Leppanen
77
171
(female)
15(8.8%)** Drop
Jump

NC ACL
tear
3 years
• â hip ROM
• á peak knee
extensor moment

Leppanen
79
171
(female)
15
(8.8%)**
Drop
Jump

NC ACL
tear
3 years
• â peak knee
flexion
• á peak vGRF

Leppanen
80

258 (111
female &
147
male)
8 (3.1%)**
female
Standing
knee lift
NC ACL
tear
12 months
• á pelvic tilt
(females only)
• previous ACL tear

* All predictors 3D unless otherwise indicated. Strongest predictor in bold, if available in multivariate
model.
** Included athletes with previous ACL injuries.
Abbreviations. NC: non-contact; ROM: range of motion; IC: initial contact; vGRF: vertical ground
reaction force

6
Consistent with the findings for initial injury risk, biomechanical prediction
studies for secondary ACL injury are conflicting. For example, Paterno et al. reported
that increased knee valgus motion was a predictor of secondary ACL injury during a
drop jump (albeit when combined with other factors),
110
while studies by King et al.
reported that increased knee valgus angles/moments were not predictors during any of
the 4 tested tasks (drop jump, single leg drop jump, planned cut, unplanned cut).
70,71

Differences in study methodology may be partly responsible for the disparate findings.
For example, Paterno et al. evaluated female athletes, while King et al. studied males.
It is well established that males and females differ in valgus motion (with females
displaying higher values),
15,127,137
which makes it difficult to compare studies with
different sex distributions.
Moreover, given that increased frontal plane motion at the knee (2D or 3D
valgus) has been shown to predict initial
59
and secondary ACL injury during a drop
jump,
110
the results of Poston et al.
119
that decreased frontal plane motion (summed at
the trunk, pelvis, and knee) during a single-leg land from a box predicts ACL re-injury
are equally confusing. Decreased frontal plane motion (or a rigid posture) as defined
by Poston et al. can result from decreased ipsilateral trunk lean, decreased valgus,
and/or decreased pelvis drop. As such, a varus bias (rather than a valgus bias) was more
likely to be a risk factor.
A reduction in movement in the frontal plane (which has been shown to increase
injury risk) may be occurring in tandem with a reduction of motion in the sagittal plane
posture.
119
Although decreased motion in the sagittal plane (defined by decreased hip
range of motion and decreased knee flexion) increases initial ACL injury risk in
females,
77,78
the opposite has been reported for secondary ACL injury (albeit in
males).
70,71
It is important to note that although King et al. reported that a softer sagittal
posture predicted secondary ACL injury during a drop jump (which included increased
knee flexion for ipsilateral tears and decreased height of the COM for contralateral),
70,71

7
this was combined with other movement risk factors, which again highlights the
complexity in interpreting ACL risk factors.

Table 2-2. Movement Risk Factors for Secondary ACL Injury

Author N Injured
(%)
Task Endpoint Follow
-up
Time
Significant Risk Factors*
Paterno
110
56 (25
female
& 21
male)
13 (23%)

3 ipsil,
10 contra
Drop
Jump

Ipsil or
contra
NC ACL
tear
12 mo
• â net hip external rotator
moment impulse
• á knee extensor moment
asymmetries
• á 2D knee valgus
• á postural instability

Poston
119

49 (32
female
& 17
male)
7
(14%)**
Single-leg
land from
box
Ipsil NC
ACL Tear
24 mo
• â summation 2D knee
valgus, 2D trunk lean, 2D
pelvis tilt (rigid posture)
• â summation 2D valgus,
2D trunk lean (rigid
posture)

King
70

1045
males
38
(3.6%)
Drop
jump
&
unplanned
90° cut

Ipsil NC
ACL Tear
24 mo Drop Jump:
• á knee flexion
• â vertical distance COM
to ankle
• á ground contact time
Unplanned Cut:
• á asymmetry trunk lean
• á asymmetry pelvic drop
• á asymmetry distance
COM to knee

King
71

1045
males
67
(6.4%)
Drop
jump
& single
leg drop
jump

Contra
NC ACL
Tear
24 mo Drop Jump:
• á hip extension moment
• á hip external rotation
moment
• â vertical distance COM
to ankle
• á ground contact time

Single leg drop jump:
• â vertical distance COM
to ankle
• â COM stiffness
• á vGRF early stance
• â vGRF midstance
• á hip extension moment

* All predictors 3D unless otherwise indicated. Strongest predictor in bold, if available in multivariate
model.
** Included injuries beyond 2
nd
ACL tear
Abbreviations. NC: non-contact; COM: center of mass; vGRF: vertical ground reaction force.

8
Summary
Comparisons among studies that have attempted to identify movement risk
factors for ACL injury (initial or secondary) are difficult owing to highly varied
approaches in study design and statistical approaches. For example, some studies have
identified movement risk factors after controlling for non-movement factors that may
influence ACL injury,
77,78,80
while other studies have combined multiple movement risk
factors into a single model (by creating novel metrics or entering them separately).
70,71,119
In addition, with regards to secondary ACL injury, surgical factors influence
injury risk, but many studies fail to control for non-movement risk factors,
110,119
which
makes it difficult to ascertain the true influence of movement. Furthermore, study
designs vary widely with how movement risk factors are defined. For example, the
vGRF has been explored as a risk factor using the peak value, or average value across
different timeframes.
71,79
Lastly, studies are difficult to compare owing to differences in
how movement is assessed (2D, 3D, or 2D/3D), the sex of the study sample (females,
males, or males/females), definition of ACL injury (initial, initial or subsequent,
ipsilateral, contralateral, or ipsilateral/contralateral), variations in task, and follow-up
time (ranging from 12 months to 7 years).
59,70,71,75,77,79,80,101,110,116,119

Currently Available Movement Assessments
Impaired movement has been implicated in increased risk of injury at the low-
back, hip, knee, lower leg, and ankle.
25,36,59,64,70,71,77,79,80,92,96,110,119,125
Therefore, it is not
surprising that more than a dozen clinical movement assessments have been
documented in the literature.
20,27,38,53,86,87,89,97,99,100,105,106,123,146
Despite a sizeable amount of
literature on various movement assessments, none has gained adequate supporting
evidence to justify usage for the purposes of injury prediction.
86,88,94,104,140

With regard to ACL injury, the LESS movement screen was developed to
identify high-risk movement patterns during a double-limb drop.
105
However, the ability
of LESS scores to predict initial ACL injury is limited to youth soccer, with no

9
predictive capacity in high-school and collegiate athletes.
104,140
A shortcoming of LESS
is the fact that only a double limb landing task is evaluated. This limitation is notable
as ACL injuries typically occur during single-limb, multidirectional tasks.
4,29
Although
movement impairments during a drop jump are predictive of ACL injury in a selective
number of studies (based on 3D motion analysis),
59,70,71,77,79,110
this has not been a
consistent finding among studies (based on 2D angular measures or 3D motion
analysis).
75,101,116

Currently available movement assessments are limited by 3 fundamental issues.
First, many movement assessments include evaluation of a single task.
27,38,97,99,100,105
This
is problematic as a single task assessment does not capture the wide range of movements
that may be contributory to injury. A thorough assessment should evaluate multiple
tasks across a wide range of sport specific demands. Second, many movement
assessments do not evaluate tasks that contribute to most sport injuries, which includes
impact with the ground (i.e., landing), sudden deceleration maneuvers, and/or change
in direction (i.e., pivoting or cutting).
4,5,29,162
Lastly, it is important that a comprehensive
movement assessment evaluate the specific movement attributes that have been shown
to predict future injury. For example, prospective studies highlight the importance of
assessing movement behavior in the frontal and sagittal planes as multiplanar movement
impairments have been linked to both primary and secondary ACL
injury.
59,70,71,77,79,80,110,119

Introducing the Movement Performance Assessment (MPA)
Given limitations of current movement assessments described in the literature,
the MPA was created based on 2 fundamental principles: (1) to examine movement
quality in the frontal and sagittal planes across a wide range of sport-specific tasks
(single limb and double limb), and (2) to comprehensively quantify movement
impairments at the trunk, pelvis, hip, and knee. The MPA utilizes inexpensive, 2D
video analysis technology to examine movement quality across a wide range of sport-

10
specific tasks (single limb and double limb), including those that involve a change of
direction. The evaluation criteria were designed to identify movement impairments at
the trunk, pelvis, hip, and knee, which have been hypothesized to be contributory to
ACL injury.
59,70,71,77,79,80,110,119
In addition, the MPA examines movement quality across
6 tasks that challenge the athlete across multiple domains (such as deceleration, single
leg landing, change of direction) that are thought to underlie ACL injury. An overview
of the MPA is provided in Table 2-3.

MPA Movement Domains
Knee strategy provides an estimate of the
hip to knee extensor moment ratio by measuring
the 2D trunk-tibia angle at peak knee flexion
(Figure 2-1). The trunk-tibia angle is calculated
using the difference between the 2D trunk and
tibia inclination angles. Trunk inclination is
measured as the angle between a vertical line
Table 2-3. Overview of Movement Performance Assessment (MPA)

Task Dominant
Plane of
Motion
Single vs.
Double
Limb
Incorporates
Change of
Direction
Sport Specific Task
Where ACL Injury
Occurs

Movement Domains
Assessed
Step Down Sagittal Single No No Pelvis stability
Trunk stability
Knee stability
Knee strategy
Drop Jump Sagittal Double No Yes Knee stability
Knee strategy
Shock absorption
Lateral
Shuffle

Frontal Single Yes Yes Pelvis stability
Trunk stability
Knee stability
Knee strategy
Shock absorption
Deceleration

Sagittal Single Yes Yes
Triple Hop

Sagittal Single No Yes
Side-Step-Cut Multiplanar

Single Yes Yes
Figure 2-1. Knee strategy is scored using the
trunk-tibia angle (+trunk more forward
than tibia, -tibia more forward than trunk).
Scored during all 6 tasks.

11
starting at the hip joint center and a line bisecting the torso. Tibia inclination is
measured as the angle between a vertical line starting at the ankle joint center and a
line through the ankle and knee joint centers. A lower trunk angle relative to tibia angle
(i.e., smaller trunk-tibia angle) is indicative of decreased use of the hip extensors over
the knee extensors during the lowering phase of squatting.
8
In addition, increased use
of the knee extensors (moments) during landing predicts initial non-contact ACL injury
(Table 2-1).
77

Shock absorption provides an estimate of the
peak vGRF by measuring the 2D thigh angle at
peak knee flexion (Figure 2-2). The thigh angle is
measured as the angle between a line bisecting the
knee (perpendicular to the ground) and a line
approximating the bisection of the femur.
Decreased ability to attenuate the force of impact
(i.e., increased peak vGRF) predicts initial ACL
injury (Table 2-1).
79
High vGRFs place the ACL at high risk for injury, as ACL rupture
occurs within 17-50 ms after initial contact with the ground,
74
which corresponds to
the time frame to the peak vGRF.
73

Pelvis stability provides an estimate of
pelvis tilt by measuring 2D pelvis tilt at peak
knee flexion (Figure 2-3). Pelvis tilt is measured
as the angle between the line connecting the
ASIS’s and a horizontal line starting at the ASIS
of the stance limb. Increased pelvis tilt during a
standing knee lift predicts initial non-contact
ACL injury (Table 2-1).
80

Figure 2-2. Shock absorption is scored
using the thigh angle (+increased hip
and/or knee flexion). Scored during 5 of 6
tasks (all except step down).
Figure 2-3. Pelvis stability is scored using
pelvis tilt (+drop, -rise). Scored during 5 of
6 tasks (all except drop jump).

12
Trunk stability provides an estimate of trunk
lean by measuring 2D trunk lean at peak knee
flexion (Figure 2-4). Trunk lean is measured as
the angle between a vertical line starting at the
umbilicus and a line through the umbilicus and
sternum. Lateral trunk lean has been reported to
predict ACL re-injury when combined with other
frontal plane measures (i.e., FPPA and/or pelvis
tilt) during a single leg land (Table 2-2).
119

Knee stability provides an estimate of
knee valgus angles/moments by measuring the
2D frontal plane projection angle (FPPA) at
peak knee flexion (Figure 2-5). The FPPA is
measured as the angle formed by 3 points
(ASIS, knee joint center, ankle joint center).
This value is subtracted from 180 to represent
the anatomical frontal plane alignment of the
knee. An increase in the FPPA (i.e., valgus) is associated with increased frontal and
transverse kinematics at the hip and knee,
130,131,138,158
which are risk factors for ACL
injury (Table 2-2).
59,110

Summary
The MPA was developed to fill a much-needed gap in the sports medicine
literature. It is proposed that the combination of single-limb and double-limb
movements, as well as tasks that involve a change in direction, is more representative
of the demands required by the athlete participating in high-risk sports. Based on
previously published prospective studies related to primary and secondary ACL injury
risk, the MPA broadens the scope of what a movement scoring system should entail,
Figure 2-5. Knee stability is scored using the
frontal plane projection angle (+valgus, -
varus). Scored during all 6 tasks.

Figure 2-4. Trunk stability is scored using
trunk lean (+ipsilateral lean, -contralateral
lean). Scored during 5 of 6 tasks (all
except drop jump).

13
thus establishing a critical bridge between clinical practice and research. Although the
MPA has been developed, it remains unknown if the 2D measures represent/predict
3D measures. In addition, whether the MPA movement domains are relevant for
predicting ACL re-injury remains uncertain. This dissertation will address these
questions as a first step in establishing the MPA as a potential clinical tool to assess an
athlete’s readiness to return to sport following ACLR.

14
CHAPTER 3:
CLINICAL ESTIMATION OF THE USE OF THE HIP AND KNEE EXTENSORS
DURING ATHLETIC MOVEMENTS USING 2D VIDEO

Published in J Appl Biomech. 2021; 37(5): 458-462.

Introduction
Although sport participation is linked to a multitude of health benefits,
41,42
the
risk of injury is of particular concern. Studies have cited the knee as the most commonly
injured body region, accounting for as high as 59% and 42% of all injuries for females
and males, respectively.
32,67
Identified causes of knee injury in sport are many and
include diminished lower-extremity strength,
69,128
poor balance and/or
proprioception,
35,39,110,117
and movement behaviors that increase knee loading.
36,59,77,78,110

With respect to movement behavior and knee injury, performing athletic tasks with
elevated knee extensor moments increases load on the patellar tendon,
40
patellofemoral
joint,
149
and the anterior cruciate ligament (ACL).
16,132
Additionally, high knee extensor
moments have been found to predict ACL injury.
77,110

Females have been reported to favor use of the knee extensors over the hip
extensors to decelerate the body center of mass during landing.
141
More specifically,
females exhibit higher knee extensor moments relative to hip extensor moments during
landing when compared to males.
141
This is important, as athletes who land with greater
knee extensor moments relative to hip extensor moments display greater knee valgus
moments and angles,
118
both of which have been shown to predict ACL injury.
59,110

Although the hip/knee extensor moment ratio (HKR) can be used to characterize
movement behavior that is indicative of greater use of the knee extensors relative to
the hip extensors, quantification of hip and knee extensor moments requires expensive
and sophisticated laboratory equipment (i.e., force plates and 3D motion analysis).

15
Characteristics of movement behavior associated with higher knee extensor
moments (or quadriceps muscle activity) during athletic tasks include an upright
trunk
11,12,142
and forward displacement of the tibia (knee pass the toe).
46,68,84,142
In a
previous publication, our group reported that the relative orientation of the trunk and
tibia in the sagittal plane could be used to estimate the average HKR during various
squat conditions.
7
More specifically, the difference between the sagittal plane trunk and
tibia inclination at peak knee flexion explained 70% of the variance in the average
HKR during the lowering phase of the squat. However, it is not clear whether this
method can be used to evaluate tasks that involve impact with the ground (i.e., single
or double limb landing) or a change in direction (i.e., cutting).
Given that identification of movement strategies that may contribute to knee
injury is important from a clinical standpoint, there is a need for a practical method to
characterize movement behavior that is indicative of how individuals utilize the hip
and knee extensors during dynamic tasks. The current study represents an extension of
our previous work, by evaluating whether the difference between sagittal plane trunk
and tibia orientations obtained from 2D video (2D trunk-tibia) could be used to predict
the average HKR during sport-specific activities. We hypothesized that the 2D trunk-
tibia angle would be predictive of the average HKR during the deceleration phase of a
wide range of athletic movements.

Methods
Participants
Thirty-nine healthy athletes (15 males and 24 females) between the ages of 13
and 40 years participated in this study (15 males and 24 females; Table 3-1). All
participants were currently partaking in a sport with high levels of jumping, cutting, or
lateral movements (such as soccer, basketball, volleyball, lacrosse, football, netball, or
tennis). Participants were excluded if they had current lower-extremity pain, any
history of ACL reconstruction, lower-extremity injuries/surgeries in the past 6 months,

16
or indicated any medical condition that would impair their ability to perform the
athletic tasks.

Sample-Size Calculation
A sample size calculation was performed based on pilot data to determine the
number of participants needed to assess the relationship between the 2D trunk-tibia
angle and average HKR across the 6 tasks. Using a 5% significance level, 90% power,
R
2
value of 0.46, and 1 predictor, a minimum of 15 participants was deemed necessary.

Instrumentation
Ground reaction forces were collected at 1200 Hz (Model #BP600900-2000,
Advanced Mechanical Technology, Inc, Watertown, MA, USA) and synchronized with
the motion capture system. The force plate was embedded into the floor and was used
for 5 out of the 6 tasks evaluated. For the step-down task described below, a portable
force plate was integrated into a 22 cm step (Model #O60-7000, Advanced Mechanical
Technology, Inc, Watertown, MA, USA).
Two-dimensional kinematic data were collected at 120 Hz using a video-based
8-camera motion analysis system (Simi Reality Motion Systems GmbH,
Unterschleissheim, Germany). One of the 8 cameras was positioned 80 cm off the
ground (perpendicular to the force plate) and was used to collect the required sagittal
plane images for the 2D analysis.

Table 3-1. Characteristics of Study Participants, Mean (SD)

Males (n = 15)

Females (n = 24)
Age (yrs)

23.8 (7.3) 17.3 (6.3)
Height (m)

1.8 (0.1) 1.7 (0.1)
Mass (kg)

78.9 (16.2) 56.1 (11.3)
BMI (kg/m
2
)

23.6 (4.2) 20.2 (2.9)

17
Procedures
Prior to data collection, participants were informed about the nature of the study
and written consent was obtained as approved by the Institutional Review Board of the
Health Sciences Campus at the University of Southern California (Los Angeles, USA).
Once informed consent was obtained, participants warmed up on a stationary bike for
5-10 minutes.
Participants were instrumented with 21 reflective markers (10 mm diameter) on
the right lower extremity. Semi-rigid plastic plates with mounted markers were used
for the thigh, tibia, and heel clusters. In addition, markers were placed on the following
bony landmarks: distal aspect of the 2nd toe, 1st and 5th metatarsal heads, medial and
lateral malleoli, medial and lateral femoral epicondyles, bilateral greater trochanters,
bilateral iliac crests, and bilateral anterior superior iliac spines. For the torso, markers
were placed on the L5-S1 junction, C7, sternal notch, and acromioclavicular joints
(bilateral). A standing static calibration trial was obtained to determine the local
segment coordinate system and joint axes. The ankle markers (medial and lateral
malleolus), knee markers (medial and lateral epicondyles), toe markers (distal aspect
of the 2nd toe, 1st and 5th metatarsal heads), greater trochanters, and anterior superior
iliac spines were removed prior to the dynamic trials.
Two-dimensional video were collected during the following tasks: 1) Step Down,
2) Drop Jump, 3) Lateral Shuffle, 4) Deceleration, 5) Triple Hop, and 6) Side-Step-
Cut. Details regarding the instructions provided to participants for each of the tasks
can found in Table 3-2. A trial was considered successful if all markers remained visible
and only the foot of tested limb fully contacted the force plate. Participants were
permitted to practice until comfortable with the performance of each task. All tasks
were performed at a self-selected speed, and 1 to 2 trials were obtained for each of the
tasks above.

18

Data Analysis
The first successful trial was selected for each task and used for data analysis.
Marker position data were labeled in Simi Motion and then exported with the force
data to Visual3D software (C-Motion, Inc, Germantown, MD, USA). Marker
trajectory and analog force plate data were low-pass filtered at 12 Hz, using a fourth-
order Butterworth filter.
150
Joint angles were calculated using a X-Y-Z (sagittal-frontal-
transverse) Cardan sequence. The trunk was modeled as a single rigid segment, defined
proximally by 2 iliac crest markers and distally by 2 acromion markers.
Inverse dynamics equations were used to calculate net joint moments (internal)
at the hip and knee. Moment data were normalized to body mass. The average hip and
knee extensor moments during the deceleration phase of all tasks (initial contact to
peak knee flexion) were calculated. For the step down, the average hip and knee
extensor moments were calculated during the lowering phase (initiation of the
movement to the time at which the heel touched the ground). The dependent variable
Table 3-2. Description of Athletic Tasks

Tasks Description

Step Down Participants were instructed to lower themselves from a 0.22 m step, tap the opposite
heel to the floor, then return to the starting position. This motion was repeated 5
times without stopping.

Drop Jump Participants stood on a 0.46 m box and were instructed to drop from the box, land
with only the tested limb on the force plate, then jump as high as possible.

Lateral Shuffle Participants were instructed to shuffle to the side as quickly as possible (4.6 m
runway), plant only the tested limb on the force plate, then switch directions and
shuffle back to the start. This motion was repeated 2 times without stopping.

Deceleration Participants were instructed to run forward as quickly as possible (4.6 m runway),
plant only the tested limb on the force plate, then backpedal to the starting position.
This motion was repeated 2 times without stopping.

Triple Hop Participants were instructed to perform 3 consecutive maximal forward hops on the
tested limb and stick the landing on the force plate. The starting distance was 90%
of the maximal hop length, measured from the center of the force plate. Maximal
hop length was established prior to biomechanical testing.

Side-Step-Cut Participants were instructed to run forward as quickly as possible (4.6 m runway),
plant only the tested limb on the force plate, then turn 90°.

19
of interest was the average HKR, which was calculated by dividing the average hip
extensor moment by the average knee extensor moment for each task.
For the 2D video analysis, the image containing peak knee flexion was identified
visually. For the step down, the frame at which the contralateral heel touched the
ground was used for analysis. Images were uploaded into a free image processing
software for 2D angle assessments (ImageJ, Version 1.50i, National Institute of Health,
USA). Trunk inclination was measured as the angle between a vertical line starting at
the hip joint center and a line bisecting the torso (Figure 2-1). A positive value
represented a forward trunk position while a negative value represented backward trunk
lean. Tibia inclination was measured as the angle between a vertical line starting at the
ankle joint center and a line through the ankle and knee joint centers (Figure 2-1). A
positive value represented a forward tibia (i.e., dorsiflexion), while a negative value
represented a backward inclination (i.e., plantarflexion). The difference between the
2D trunk and tibia inclination angles was calculated for each trial and used for statistical
analysis. All 2D trunk and tibia measurements were obtained by a single investigator
who demonstrated excellent intra-rater reliability for all tasks prior to the start of the
study (intraclass correlation coefficients (ICCs) ranging from 0.95 to 1.0).

Statistical Analysis
Linear regression analysis was used to assess the ability of the 2D trunk-tibia
angle (independent variable) to predict the average 3D HKR (dependent variable). This
analysis was repeated for each task. All analyses were evaluated for a 2D trunk-tibia
angle*sex interaction. All regression equations were adjusted for body mass.
Each analysis was screened for outliers using standardized residuals; absolute
values greater than 3 were deemed outliers and removed. All statistical analyses were
performed using SPSS (Chicago, Illinois, USA) and a custom MATLAB script (The
Mathworks, Inc., Natick, MA) with alpha set at 0.05.

20
Results
Due to force plate malfunction, force (moment) data was not available for one
subject during the drop jump (1 male) and 8 participants (6 females) during the step-
down. Descriptive statistics for the average HKR and 2D trunk-tibia inclination angles
for each task are presented in Table 3-3. Of the 6 separate linear regression analyses
performed, one outlier for the deceleration task was identified and removed.

There were no 2D trunk-tibia angle*sex interactions for any of the linear
regression analyses performed. As such, the data were collapsed across males and
females. Each of the regression models were found to be significant when adjusted for
body mass (all p < 0.013). More specifically, the R
2
values were as follows: step down
(0.77), triple hop (0.55), deceleration (0.41), drop jump (0.30), shuffle (0.30), and side-
step-cut (0.17). The beta coefficient for the 2D trunk-tibia angle for all models was
positive, indicating that an increase in the 2D trunk-tibia angle at peak knee flexion
significantly predicted an increase in the average HKR (Figure 3-1).
Table 3-3. Average Hip/Knee Extensor Moment and 2D Trunk-Tibia Angles at Peak Knee Flexion

Step Down Drop Jump Lateral
Shuffle

Deceleration Triple Hop Side-Step-
Cut

Average
HKR

0.66 (0.36) 0.93 (0.33) 0.75 (0.29) 1.14 (0.58) 0.67 (0.25) 0.84 (0.34)
2D Trunk-
Tibia (deg)

-23.9 (16.2)

3.1 (13.6) 5.5 (11.5) 13.7 (17.8) 2.8 (14.0) 3.7 (15.3)
Abbreviations. HKR: hip/knee extensor moment ratio. Data presented as mean (SD).

21

Discussion
The current study sought to determine whether the 2D trunk-tibia angle could
be used to approximate the relative demand of the hip and knee extensors (average
HKR) during athletic tasks. Consistent with our hypothesis, the 2D trunk-tibia angle
was able to predict the average HKR for a wide range of athletic movements, with
higher angles predicting higher average HKRs (Figures 3-1). Specifically, results of the
linear regression analysis indicated that 17% to 77% of the variance in the average
HKR could be explained by the 2D trunk-tibia angle at peak knee flexion across the
tasks evaluated.
Figure 3-1. Linear regression models to predict the average hip/knee extensor moment for
each task based on the 2D trunk-tibia co-varied for body mass. Predicted 3D average HKR
values were calculated by regressing the 2D trunk-tibia and body mass on the observed
average HKR. The beta coefficient for the 2D trunk-tibia for all models was positive.
Abbreviations. HKR: hip/knee extensor moment ratio.

22
The step-down task exhibited the strongest linear relationship with an R
2
of 77%
(Figure 3-1). This finding is consistent with Barrack et al.
7
who reported that the trunk-
tibia angle at peak knee flexion (using 3D video) during bilateral squatting explained
70% of the variance in the average HKR. In contrast, the 2D trunk-tibia angle during
the activities that were performed more quickly and required a change in direction of
the body center of mass demonstrated a lower predictive capacity for the average HKR
(R
2
values ranging from 17% to 55%) (Figure 3-1). The fact that moments are
influenced by ground reaction forces resulting from the acceleration of the body center
of mass suggests that the static 2D trunk-tibia angle measurement was less capable of
predicting the HKR during the more dynamic tasks. In addition, the 2D trunk-tibia
angle is prone to error with out of plane movements, which may explain the high
variability observed for the 2D trunk-tibia measures (Table 3-3) and why side-step-
cutting had the lowest predictive ability (R
2
= 17%).
The results of the current study suggest that the 2D trunk-tibia angle
measurement may be useful as a method to screen individuals at risk for knee injury.
Other screening methods using 2D video have been developed to predict knee injury,
most notably the Landing Error Scoring System (LESS). The LESS scores kinematic
attributes during a double-limb drop jump that are thought to contribute to injury.
105

In contrast to assessing 2D kinematic errors, the method described in the current study
utilized 2D video to gain insight into kinetic factors (i.e., moments) that may contribute
to injury. Although the ability of LESS to predict injury has been reported to be
mixed,
104,140
movement screens that incorporate attributes of kinematic and kinetic risk
factors may provide greater injury predictive value.
There are several limitations within the current study that warrant discussion.
First, our data was derived using healthy individuals. As such, our results may not be
applicable to injured populations. Second, only the deceleration or lowering phase of
each task was considered in our analysis. Our results may not apply to the acceleration
phase of the tasks evaluated. Third, the 2D trunk-tibia variable is prone to out of plane

23
measurement error. As such, the ability of the 2D trunk-tibia angle to predict the
average HKR should be interpreted with caution during activities that incorporate large
amounts of out-of-plane movement (i.e., cutting). Lastly, we only analyzed data from
a single trial. Whether or not the level of predictability reported here would improve
with the averaging of data from multiple trials remains to be seen.

Conclusion
The findings of the current study suggest that 2D video analysis can be used to
predict the average HKR, across a wide range of athletic tasks. As such, the 2D trunk-
tibia angle can be used as a practical method to characterize movement behavior that
is indicative of how individuals utilize the hip and knee extensors during dynamic tasks.
Further investigation is needed to determine if the 2D trunk-tibia, as described in the
current study, can predict knee injury (in isolation or combined with established
movement screens).

24
CHAPTER 4:
ESTIMATION OF VERTICAL GROUND REACTION FORCE PARAMETERS
DURING ATHLETIC TASKS USING 2D VIDEO

Published in Gait Posture. 2021; 90: 483-488.

Introduction
Although sport participation is linked to a multitude of health benefits,
41,42
knee
injuries are common among those who participate in athletics and can negatively
impact sport performance. Ligament damage is the most common knee pathology, with
the anterior cruciate ligament (ACL) being one of the most frequently injured
ligamentous structures.
23
Sports that require frequent jumping, landing, and cutting,
(e.g., soccer, volleyball, and basketball) are among the sports with the highest ACL
injury rates.
2,65
Although the proposed causes of ACL injury are many, high impact
forces during landing have been reported to be contributory.
78

The relationships among various measures of ground reaction forces (GRF) and
ACL injury have been explored in several studies. Based on a prospective study, a
higher peak vertical ground reaction force (vGRF) during double-limb landing has
been reported to be predictive of future ACL injury.
78
vGRF impulse also has been
shown to be important relative to ACL injury risk as cadaveric simulations have shown
that applying a simulated vGRF impulse during landing, with the addition of an
external knee valgus moment, anterior shear force, and internal tibial rotation, induces
ACL failure.
9
In addition, greater vGRF impulse asymmetry has been reported to
predict knee extensor moment asymmetry during side-cutting and stop-jumping in
persons who have undergone ACL reconstruction.
28
This is important as a prospective
study has reported that knee extensor moment asymmetry has been shown to be
predictive of ACL re-injury.
110

25
During athletic tasks, the hip and knee extensors are important muscle groups
responsible for attenuating impact forces during landing.
34,118
Performing landing
activities with high degrees of leg stiffness (i.e., decreased hip and knee flexion) results
in elevated vGRFs.
34,91,139,151
In turn, increasing hip and knee flexion during landing has
been reported to result in lower vGRFs and lower knee valgus angles and moments.
118

In addition, studies employing musculoskeletal modeling simulations in healthy persons
and those who have undergone ACL reconstruction have shown that increasing hip
and knee flexion during landing results in decreased tibiofemoral joint compressive
forces
151,152
and decreased ACL tensile forces.
76

Given the importance of elevated vGRFs in contributing to ACL injury, there
is a need for a clinic-friendly method to characterize movement behavior that is
representative of elevated impact forces. In the current paper, we propose that the
sagittal plane thigh angle, as quantified with 2D video analysis, may be a viable method
to assess movement behavior that is predictive of vGRFs. Given that the thigh is the
connecting segment between the hip and knee, it is possible that a single measurement
that reflects flexion of these joints may provide insight into how an individual attenuates
impact forces.
The purpose of the current study was to determine whether images obtained
from 2D sagittal plane video could be used to predict measures of vGRFs during a
wide range of athletic tasks. More specifically, we sought to determine whether the 2D
thigh angle obtained at peak knee flexion could be used to predict the peak vGRF and
vGRF impulse during single limb and double limb landings and movements that involve
a change of direction. We hypothesized that an increased 2D thigh angle (which
corresponds to increased hip and knee flexion) would be predictive of lower peak
vGRFs and/or lower vGRF impulse. Information gained from this study will provide
clinicians with a practical method to characterize movement behavior associated with
high impact forces.

26
Methods
Participants
Thirty-nine healthy athletes between the ages of 13 and 40 years partook in this
study (15 males and 24 females; Table 3-1). All participants were currently partaking in
a sport with high levels of jumping, cutting, or lateral movements (such as soccer,
basketball, volleyball, lacrosse, football, netball, or tennis). Participants were excluded
if they had current lower-extremity pain, any history of ACL reconstruction, lower-
extremity injuries/surgeries in the past 6 months, or indicated any medical condition
that would impair their ability to perform the athletic tasks.

Sample-Size Calculation
A sample size calculation was performed based on pilot data in G*Power
(Version 3.1) to determine the number of participants needed to assess the relationship
between the 2D thigh angle and the peak vGRF across the 5 tasks. Using a 5%
significance level, 90% power, R
2
value of 0.39, and 1 predictor, a minimum of 19
participants was deemed necessary.

Instrumentation
Two-dimensional kinematic data were collected at 120 Hz using a video-based
8-camera motion analysis system (Simi Reality Motion Systems GmbH,
Unterschleissheim, Germany). One of the 8 cameras was positioned 80 cm off the
ground (perpendicular to the force plate) and was used to collect the required sagittal
plane images for the 2D analysis. GRFs were collected at 1200 Hz and synchronized
with the motion capture system (Model #BP600900-2000, Advanced Mechanical
Technology, Inc, Watertown, MA, USA).

27
Procedures
Prior to data collection, participants were informed about the nature of the study
and written consent was obtained as approved by the Institutional Review Board of the
Health Sciences Campus at the University of Southern California (Los Angeles, CA).
Once informed consent was obtained, participants warmed up on a stationary bike for
5-10 minutes. For all data procedures outlined below, data were obtained on the right
limb.
Prior to data collection, maximum triple hop distance was obtained. Two-
dimensional video was then obtained during the following tasks in the following order:
1) Drop Jump, 2) Lateral Shuffle, 3) Deceleration, 4) Triple Hop, and 5) Side-Step-
Cut. Details regarding the instructions provided to participants for each of the tasks
can found in Table 3-1. These tasks were selected based on current knowledge of
movements thought to be associated with ACL injury. Since most ACL tears occur
during a non-contact episode when landing from a jump, decelerating, or cutting with
a change in direction,
14,60,124
we deemed it important to incorporate multiple dynamic
movements. A trial was considered successful if only the foot of the tested limb fully
contacted the force plate. Participants were permitted to practice until comfortable
with the performance of each task and could rest between trials as needed. One to two
trials were obtained for each of the tasks above.

Data Analysis
The first successful trial was selected for each task and used for data analysis.
Analog force plate data collected in Simi Motion were exported to Visual3D software
(C-Motion, Inc, Germantown, MD, USA). Analog force plate data were low-pass
filtered at 12 Hz, using a fourth-order Butterworth filter.
150

The dependent variables of interests were the first peak of the vGRF and the
vGRF impulse. The first peak was selected as this was the most consistent time point
across subjects within a given task. In addition, several of the tasks evaluated only had

28
a single vGRF peak. The first peak was examined for movements that demonstrated
two peaks to provide consistency across tasks.
Peak vGRF was obtained from the time-normalized data, while the vGRF
impulse was quantified as the integral (trapezoid rule) of the vGRF from initial contact
to the first peak vGRF using non-time normalized data. All vGRF variables were
normalized to body mass. Given the fact that movement speed was not standardized
across participants for the tasks evaluated, vGRF variables also were normalized to
approach velocity except for the drop jump as the platform height was standardized
across participants. Approach velocity was calculated by tracking a marker placed over
L5-S1.
For the 2D video analysis, the image containing peak knee flexion was visually
identified. Images were uploaded into ImageJ software (Version 1.50i, National
Institute of Health, USA) for 2D thigh angle assessments. The thigh angle was
measured as the angle between a line bisecting the knee (perpendicular to the ground)
and a line approximating the bisection of the femur (Figure 2-2). Increased 2D thigh
values were representative of increased hip and/or knee flexion. All 2D thigh
measurements were obtained by a single investigator who demonstrated excellent intra-
rater reliability for all tasks prior to the start of the study (ICCs ranging from 0.89 to
0.99).

Statistical Analysis
Linear regression analysis was used to assess the ability of the 2D thigh angle
(independent variable) to predict the vGRF (dependent variable). This analysis was
repeated for each task and was run separately for each dependent variable (peak vGRF
and vGRF impulse). R
2
values >= 0.12, R
2
>= 0.35, and R
2
>= 0.50 were considered
small, medium, and large effect sizes, respectively.
18,19
All analyses were evaluated for
a 2D thigh angle*sex interaction.

29
Each analysis was screened for outliers using standardized residuals; absolute
values greater than 3 were deemed outliers and removed. All statistical analyses were
performed using SPSS (Chicago, Illinois, USA) and a custom MATLAB script
(MathWorks, Inc, Natick, MA) with alpha set at 0.05 with alpha set at 0.05.

Results
Due to technical issues with the force plate, vGRF data was not available for 1
subject during the drop jump. Of the 5 separate linear regression analyses performed
for vGRF impulse, 1 outlier from each task (except drop jump) was identified and
removed. For the 5 separate linear regression analyses performed for vGRF, no outliers
were identified. Descriptive statistics for the vGRF variables of interest and 2D thigh
angles for each task are shown in Table 4-1. Time series data for the vGRF data for
each task is presented in Figure 4-1.

Table 4-1. Vertical Ground Reaction Force Variables and 2D Thigh Angle, Mean (SD)

Drop Jump Lateral
Shuffle

Deceleration Triple Hop Side-Step-Cut
vGRF 1
st
peak 17.0 (3.4) N/kg 7.0 (1.3)
Ns/kg-m
5.3 (1.3)
Ns/kg-m
15.1 (3.0)
Ns/kg-m
5.7 (1.1)
Ns/kg-m

vGRF impulse
to 1
st
peak
0.57 (0.14)
Ns/kg
0.29 (0.11)
Ns
2
/kg-m
0.16 (0.09)
Ns
2
/kg-m
0.38 (0.12)
Ns
2
/kg-m
0.15 (0.05)
Ns
2
/kg-m

2D thigh angle
(deg)

56.5 (9.4) 41.5 (6.2) 64.3 (8.8) 49.3 (7.5) 43.9 (11.7)
vGRF data normalized to body mass and approach velocity for all tasks except drop jump, which was
normalized to body mass only. Abbreviations. vGRF: vertical ground reaction force.

30

There were no 2D thigh angle*sex interactions for any of the linear regression
analyses performed, indicating that the relationship between the 2D thigh angle and
vGRF was not dependent on sex. As such, the data were collapsed across males and
females. The 2D thigh angle significantly predicted the peak vGRF for drop jump (R
2

= 0.17, p = 0.009), shuffle (R
2
= 0.22, p = 0.003), deceleration (R
2
= 0.47; p < 0.001),
and triple hop (R
2
= 0.25; p = 0.001) (Figure 4-2). However, the 2D thigh did not
predict the peak vGRF for cutting (R
2
= 0.04, p = 0.25) (Figure 4-2). Regarding the
vGRF impulse, the 2D thigh angle was found to be predictive for all tasks (R
2
= 0.13
to 0.39, all p < 0.025) (Figure 4-3).
Figure 4-1. Time-normalized vGRF data for the 5 tasks evaluated. Error bars represent
1 SD. Abbreviations. vGRF: vertical ground reaction force.

31

Figure 4-2. Linear regression models for each task separately to predict vGRF 1st peak.
vGRF normalized to body mass and approach velocity for all tasks except drop jump,
which was normalized to body mass only. Abbreviations. vGRF: vertical ground reaction
force.

32

Discussion
The purpose of the current study was to determine whether the 2D thigh angle
at peak knee flexion could be used to approximate measures of vGRFs during various
athletic tasks. Our results indicate that the 2D thigh angle was able to predict peak
vGRFs for 4 of 5 tasks and vGRF impulse for all 5 tasks evaluated. In all situations,
higher 2D thigh angles (which is representative of increased hip and knee flexion) was
predictive of decreased peak vGRFs and/or lower vGRF impulse. These findings
suggest that the 2D thigh angle can be used as a practical method to characterize
movement behavior that may expose individuals to high impact forces.
Across the tasks evaluated, a wide range of predictive ability was observed. For
example, the 2D thigh angle explained 17% to 47% of the variance in the peak vGRF
(Figure 4-2) and 13% to 39% of the variance in the vGRF impulse (Figure 4-3). The
Figure 4-3. Linear regression models for each task separately to predict vGRF impulse
to 1st peak. vGRF normalized to body mass and approach velocity for all tasks except
drop jump, which was normalized to body mass only. Abbreviations. vGRF: vertical
ground reaction force.

33
highest predictability was found during the deceleration task for both vGRF variables
(vGRF: R
2
= 47%; vGRF impulse R
2
= 39%). In contrast, the side-step-cut task
exhibited the lowest R
2
value (13%) for the vGRF impulse and no predictive capability
for the peak vGRF. There may be two explanations for the limited ability of the 2D
thigh angle to predict vGRF variables during cutting. First, cutting tasks generate large
anterior-posterior and medial-lateral GRFs to redirect the body center of mass.
135
The
2D thigh angle as described in the current study was developed as a possible method
to predict vGRFs and would not necessarily be expected to predict shear forces. Given
that cutting tasks generate large shear GRFs forces relative to the vGRF, it is not
surprising that the 2D thigh angle exhibited poor predictability during this task. Second,
cutting involves a high degree of lower limb rotation.
55
Such out of plane motion would
have resulted in less accuracy in our 2D sagittal plane measurement. As such, care
should be taken in using the 2D thigh angle to estimate vGRF variables during tasks
that involve large degrees of lower-extremity rotation.
It is important to note that the 2D thigh angle as described in the current study
only was representative of the potential contribution of the hip and knee extensors with
respect to the attenuation of impact forces. Previous studies have shown that the ankle
plantarflexors also play an important role in attenuating impact forces during
landing.
31,126
For example, Rowley et al. provided real-time feedback to participants
during drop landings to produce 5 different target plantarflexion angles. The authors
reported that as plantarflexion increased at initial contact, the peak vGRF also
decreased.
126
It is likely that 2D measurements that also account for the contribution
of the ankle in attenuating impact forces may result in a better estimation of peak
vGRF and vGRF impulse.
When comparing the results between the 2 GRF variables examined, the vGRF
impulse models yielded a higher R
2
value for 3 of 5 tasks (Figure 4-3), while the peak
vGRF peak models yielded a higher R
2
value for 2 of 5 tasks (Figure 4-2). The
measurement of vGRF impulse is contingent on both force and time and may be more

34
representative of overall loading on the musculoskeletal system, compared to the peak
vGRF (which represents loading at single point in time).
28
Although the peak vGRF
during landing has been reported to be predictive of future ACL injury,
78
the results of
the current study may have clinical utility beyond the characterization of movement
behavior of persons who have experienced an ACL tear. For example, individuals with
self-reported ankle instability have been reported to exhibit higher peak vGRFs
compared to healthy controls during cutting movements.
30
Similarly, dancers with
patellar tendinopathy display greater peak vGRFs and vGRF impulse compared to
healthy controls during landing.
44
Future research is needed to determine if the 2D
thigh angle is predictive of vGRFs in persons with various clinical conditions.
The fact that an increased 2D thigh angle could predict both a decrease in the
peak vGRF and vGRF impulse suggests that this measurement may have utility in the
screening of athletes thought to be a risk for lower-extremity injury. One commonly
used method to assess lower-extremity injury risk is the Landing Error Scoring System
(LESS).
105
The LESS evaluates a wide range of metrics during a double-leg drop jump,
including the construct of shock absorption. However, the evaluation of shock
absorption in the LESS is based on subjective visual assessment of hip, knee, and ankle
flexion and overall leg stiffness during landing. The results of the current study (i.e.,
2D thigh predicts the peak vGRF and/or vGRF impulse) indicate that a simple
measurement may provide a higher level of accuracy in assessing how athletes manage
ground impact during landing. Future research should be directed towards identifying
specific 2D thigh angle cut-offs to distinguish between high-impact and low-impact
landing strategies.
Although the 2D thigh angle was predictive of the peak vGRF and/or vGRF
impulse for all tasks evaluated, the R
2
values were small to medium (0.17 to 0.47 for
peak vGRF; 0.13 to 0.39 for vGRF impulse). As such, the 2D thigh angle should not
be considered a surrogate measure of vGRF parameters. Currently it is not known what
degree of predictability of the GRF using the proposed 2D approach is important for

35
the screening of athletes. Future studies are needed to determine if the 2D thigh angle
has any predictive value in terms of lower-extremity injury.
There are several limitations of the current study that should be acknowledge.
First, we only evaluated one 2D measurement. It is possible that the addition of other
2D measurements (i.e., knee flexion or ankle plantarflexion) may have resulted in
higher predictability using a multiple regression approach. However, it was our intent
to develop a simple angular measurement that would represent the impact attenuation
ability of more than one joint. Second, only the deceleration phase of each task was
evaluated. As such, our results may not be applicable to the acceleration phase of the
tasks evaluated. It should be emphasized however, that most ACL injuries occur during
the deceleration phase of athletic movements.
14
Third, only healthy subjects were
evaluated. As such, our findings cannot be generalized to various patient populations.
As noted above, further study of the relationship between the 2D thigh angle and vGRF
measures should be performed in patient populations that may be susceptible to high
impact forces (i.e., persons post-ACL reconstruction). Lastly, we only analyzed data
from a single trial. Whether or not the level of predictability reported here would
improve with the averaging of data from multiple trials remains to be seen.

Conclusion
The findings of the current study indicate that the 2D thigh angle can be used
to approximate measures of vGRFs during various athletic tasks. An increased 2D thigh
angle (which corresponds to a softer landing strategy) was able to predict a decreased
vGRF peak for 4 of 5 tasks (all except cutting) and decreased vGRF impulse for all 5
tasks. It is proposed that the 2D thigh angle can be used as a practical method to
characterize movement behavior that may expose individuals to high impact forces.

36
CHAPTER 5:
UTILITY OF 2D VIDEO ANALYSIS FOR ASSESSING FRONTAL PLANE TRUNK
AND PELVIS MOTION DURING STEPPING, LANDING, AND CHANGE IN
DIRECTION TASKS: A VALIDITY STUDY

Published in Int J Sports Phys Ther. 2022;17(2);139-147.

Introduction
Excessive frontal plane motion of the trunk and/or pelvis has been implicated
in numerous clinical conditions.
36,64,80,125
With respect to anterior cruciate ligament
(ACL) injury, increased lateral trunk motion (with increased medial knee collapse)
during landing
36
and increased pelvis hike (i.e., frontal plane pelvis rise of the non-
stance limb) during a standing knee lift have been shown to predict ACL injury (or
non-contact knee injury of any type).
80
Regarding low back pain, increased
contralateral pelvis drop (i.e., frontal plane pelvis drop of the non-stance limb) and/or
hip adduction during single-leg landing has been shown to predict the occurrence of
symptoms in youth floorball and basketball players.
125
Based on cross-sectional studies,
greater degrees of contralateral pelvis drop and/or ipsilateral trunk lean during weight-
bearing activities have been reported in persons with patellofemoral pain (PFP)
98,157
and
groin pain
64
compared to healthy persons. Furthermore, diminished pelvis and trunk
control can contribute to increased knee valgus moments during landing,
17
a known
risk factor for the development of PFP
96
and ACL injury.
59

Due to the importance of frontal plane trunk and pelvis kinematics in
contributing to various musculoskeletal conditions, two-dimensional (2D) video
analysis has been proposed as a clinical method to identify abnormal trunk and/or
pelvis motion during dynamic tasks. To date, four studies have attempted to validate
angular measures of frontal plane pelvis and/or trunk motion obtained with 2D video
compared with three-dimensional (3D) motion with varied results.
37,72,85,131
Although

37
2D measures of pelvic drop have been shown to correlate with 3D measures during
running,
37
this finding has not been consistent across studies.
85
Similarly, studies that
have examined the validity of 2D measures of frontal plane trunk motion have reported
inconsistent findings.
72,131
While measures of 2D trunk motion have been shown to be
correlated with 3D measures during a single limb hop,
72
this finding has not been
replicated across tasks such as the single limb squat
72,131
and drop jump.
72

Apart from the diversity in the tasks analyzed, comparison of results across
studies is difficult owing to the various methods employed for defining the trunk and
pelvis segments and differences in the kinematic variables analyzed (i.e., displacement
vs. position at specific time points). In addition, previous studies in this area have not
evaluated the validity of 2D measures of pelvis and trunk motion during tasks that
involve a change in direction. This is important as high demand change of direction
tasks, such as cutting and deceleration, have been implicated in ACL injury
60,134
and
have been reported to be influenced by insufficient control at the pelvis and trunk.
36,80

To date, it is unclear whether 2D video is an appropriate surrogate for assessing
frontal plane trunk and pelvis motion as a comprehensive validity study across a wide
range of movements using a consistent methodology has not been performed. As such,
the purpose of the current study was to assess the concurrent validity and agreement
of 2D pelvis and trunk motion in the frontal plane against the respective 3D angles
during stepping, landing, and change in direction tasks. Information gained from this
study will aid in the development and/or improvement of clinical movement analysis
to identify movement impairments associated with lower-extremity injury.

Methods
Participants
Thirty-nine healthy athletes between the ages of 13 and 40 years partook in this
study (15 males and 24 females; Table 3-1). All participants were currently partaking in
a sport with high levels of jumping, cutting, or lateral movements (such as soccer,

38
basketball, volleyball, lacrosse, football, netball, or tennis). Individuals were excluded
if they had current lower-extremity pain, any history of ACL reconstruction, lower-
extremity injuries/surgeries in the past 6 months, or indicated any medical condition
that would impair their ability to perform the athletic tasks.

Instrumentation
Three-dimensional and 2D kinematic data were collected at 120 Hz using a
video-based 8-camera motion analysis system (Simi Reality Motion Systems GmbH,
Unterschleissheim, Germany). One of the 8 cameras was positioned 80 cm off the
ground (perpendicular to the force plate) and was used to collect the required frontal
plane images for the 2D analysis.

Procedures
Prior to data collection, participants were informed about the nature of the study
and written consent was obtained as approved by the Institutional Review Board of the
Health Sciences Campus at the University of Southern California (Los Angeles, USA).
Parental consent was obtained from participants younger than 18 years. Once informed
consent was obtained, participants warmed up on a stationary bike for 5-10 minutes.
Participants were instrumented with 21 reflective markers (10 mm diameter) on
the right lower extremity. Semi-rigid plastic plates with mounted markers were used
for the thigh, tibia, and heel clusters. In addition, markers were placed on the following
bony landmarks: distal aspect of the 2nd toe, 1st and 5th metatarsal heads, medial and
lateral malleoli, medial and lateral femoral epicondyles, bilateral greater trochanters
(most prominent point), bilateral iliac crests (most superior aspect), and bilateral
anterior superior iliac spines. For the torso, markers were placed on the L5-S1 junction,
C7, sternal notch, and acromioclavicular joints (bilateral). A standing static calibration
trial was obtained to determine the local segment coordinate system and joint axes.
Ankle and knee joint centers were defined as the points 50% between the malleoi and

39
femoral epicondyle markers, respectively. The hip joint centers were defined as the
points located 25% of the distance between the greater trochanter markers.
154
The ankle
markers (medial and lateral malleolus), knee markers (medial and lateral epicondyles),
toe markers (distal aspect of the 2nd toe, 1st and 5th metatarsal heads), greater
trochanters, and anterior superior iliac spines (ASIS) were removed prior to the
dynamic trials.
Two-dimensional video and 3D motion analysis were collected during the
following tasks in the following order: 1) Step Down, 2) Lateral Shuffle, 3)
Deceleration, 4) Triple Hop, and 5) Side-Step-Cut. Details regarding the instructions
provided to participants for each of the tasks can found in Table 3-1. These tasks were
selected based on current knowledge of movements thought to be associated with
various sport injuries. Participants were permitted to practice until comfortable with
the performance of each task and could rest between trials as needed. One to two trials
were obtained for each of the tasks above. As only a single repetition within a trial was
needed for statistical analysis, two trials were obtained from some tasks to ensure that
sufficient data were available in the case of technical errors (i.e., marker occlusion,
etc.).

Data Analysis
The first successful trial was selected for each task and used for data analysis. A
trial was successful if the participant performed the task as instructed with no marker
occlusion. Marker position data were labeled in Simi Motion and then exported to
Visual3D software (C-Motion, Inc, Germantown, MD, USA). Marker trajectory data
were low-pass filtered at 12 Hz, using a fourth-order Butterworth filter. Joint angles
were calculated using a X-Y-Z (sagittal-frontal-transverse) Cardan sequence. The trunk
was modeled as a single rigid segment, defined proximally by 2 iliac crest markers and
distally by 2 acromion markers. The 3D kinematic variables of interest were the frontal

40
plane trunk and pelvis angles at peak knee flexion, which were calculated relative to
the global reference frame.
For the 2D video analysis, the frame at peak knee flexion was visually identified.
For the step down, the frame at which the contralateral heel touched the ground was
used for analysis. Images were uploaded into ImageJ software (Version 1.50i, National
Institute of Health, USA) for 2D angle assessments. Pelvis tilt was measured as the
angle between the line connecting the ASIS’s and a horizontal line starting at the ASIS
of the stance limb. A positive value represented contralateral pelvis drop and a negative
represented contralateral pelvis rise (Figure 2-3). Trunk lean was measured was
measured as the angle between a vertical line starting at the umbilicus and a line
through the umbilicus and sternum. A positive value represented an ipsilateral lean
(towards stance limb), and a negative value represented a contralateral lean (away from
stance limb) (Figure 2-4). All 2D measurements were obtained by a single investigator
who demonstrated acceptable intra-rater reliability for all pelvis (ICCs ranging from
0.74 to 0.99) and trunk angles (ICCs ranging from 0.77 to 0.98).

Statistical Analysis
Data were assessed for normality using the Shapiro-Wilk’s test. Out of our 20
variables, 15 satisfied normality. Given that the majority of our data met normality and
that Pearson’s correlations are robust to extreme violations of normality,
56
all
correlations were conducted using parametric testing. Pearson’s correlation analysis was
used to assess the association between the 2D and 3D frontal plane angles at the trunk
and pelvis. Correlation coefficients were interpreted as very strong (>= 0.9), strong
(0.7-0.9), moderate (0.5-0.7), weak (0.3-0.5), and negligible (0.0-0.3).
95
Correlation
analysis was performed separately for each task and segment. Prior to analysis, all
variables were checked for outliers using Z scores. Variables with absolute Z scores >
3.0 were deemed outliers and removed.

41
Bland Altman plots were used to assess the level of agreement between the 2D
and 3D frontal plane angles at the trunk and pelvis.
13,61
Agreement was assessed
separately for each task and segment. Limits of agreement (LOA) were used to represent
the range in which an individual’s difference score fell 95% of the time, while the bias
(mean difference, MD) was used to represent the average difference between the 3D
and 2D angles (positive values indicated 3D overestimated). Prior to analysis, the MD
between 2D and 3D angles was screened for outliers using Z scores. Absolute values
greater than 3 were deemed outliers and removed. All statistical analyses were
performed using SPSS (Chicago, Illinois, USA) and a custom MATLAB script (The
MathWorks, Inc., Natick, MA) with alpha set at 0.05.

Results
Descriptive Data
Descriptive statistics for the 2D and 3D frontal plane trunk and pelvis angles are
presented in Table 5-1.

Correlation and Agreement between 2D and 3D Frontal Plane Pelvis Angles
The initial correlation analysis contained 1 outlier (deceleration), which was
removed. Pearson’s correlation analysis indicated that all 2D and 3D frontal plane
Table 5-1. Frontal Plane Angles for Pelvis and Trunk using 2D Video and 3D Motion Analysis,
Mean (SD)

Step Down Lateral
Shuffle
Deceleration Triple Hop Side-Step-Cut
2D Trunk
Lean (deg)

3.4 (6.1) -6.8 (9.0) 1.6 (5.3) 9.2 (8.6) -4.7 (10.0)
3D Trunk
Lean (deg)

3.5 (4.8) -3.8 (7.0) -1.4 (5.1) 6.6 (6.4) -8.6 (9.9)
2D Pelvis Tilt
(deg)

6.5 (4.4) 4.4 (5.2) 2.0 (4.8) -3.3 (4.7) 10.3 (8.7)
3D Pelvis Tilt
(deg)

1.9 (4.6) 7.7 (6.2) 3.0 (5.1) -2.9 (5.3) 14.5 (10.5)
Positive values for trunk lean indicate ipsilateral lean. Positive values for pelvis tilt indicate pelvic
drop.

42
pelvis angles were significantly correlated in a positive direction (r = 0.54 to 0.73, all
p < 0.001) (Figure 5-1).

The initial Bland Altman plot analysis contained 2 outliers (deceleration and
triple hop), which were removed. The mean difference (MD) between the 3D and 2D
pelvis angles ranged from -4.6° (step down) to 4.2° (side-step-cut). The 95% LOA
ranged from MD ± 7.2° (step down) to MD ± 17.1° (side-step-cut). The 95% LOA
ranged from -12.8° to 21.3° across tasks (Figure 5-2). In all tasks, the 95% LOA included
0.

Figure 5-1. Correlation models for the 2D and 3D frontal plane pelvis angles for each task.

43

Correlation and Agreement between 2D and 3D Frontal Plane Trunk Angles
The initial correlation analysis contained 1 outlier (triple hop), which was
removed. Pearson’s correlation analysis indicated that all 2D and 3D frontal plane
trunk angles were significantly correlated in a positive direction (r = 0.81 to 0.92, all p
< 0.001) (Figure 5-3).

Figure 5-2. Bland Altman plots comparing 2D vs. 3D frontal plane pelvis angles for each task. Upper
and lower dotted lines represent 95% limits of agreement. Solid line represents bias or mean difference.
Positive mean values indicate pelvis drop; negative mean values indicate pelvis rise. Abbreviations.
MD: mean difference.

44

The initial Bland Altman plot analysis contained 2 outliers (shuffle and triple
hop), which were removed. The mean difference (MD) between the 3D and 2D pelvis
angles ranged from -4° (side-step-cut) to 2.6° (shuffle). The 95% LOA ranged from
MD ± 5.5° (step down) to MD ± 7.8° (side-step-cut). The 95% LOA ranged from -
11.8° to 9.4° across tasks (Figure 5-4). In all tasks, the 95% LOA included 0.

Figure 5-3. Correlation models for the 2D and 3D frontal plane trunk angles for each task.

45

Discussion
The purpose of the current study was to assess the concurrent validity and
agreement of 2D frontal plane angles for the pelvis and trunk with the respective 3D
angles across a wide range of tasks. Our findings revealed that 2D frontal plane angles
were correlated with the corresponding 3D angles for both the trunk (strong to very
strong) and pelvis (moderate to strong). In addition, the Bland Altman plots indicated
no systematic bias, high agreement, but wide 95% LOA. These results suggest that the
use of 2D video to assess trunk and pelvis angles is appropriate, however caution is
advised when high levels of agreement or accuracy is required.
Figure 5-4. Bland Altman plots comparing 2D vs. 3D frontal plane trunk angles for each task. Upper
and lower dotted lines represent 95% limits of agreement. Solid line represents bias or mean difference.
Positive mean values indicate ipsilateral lean; negative mean values indicate contralateral lean.
Abbreviations. MD: mean difference.

46
In terms of the Pearson correlation coefficients related to the validation of 2D
trunk motion, all were strong to very strong (Figure 5-3). In addition, absolute
agreement for all tasks was below 5° and there was no systematic bias (as 0 was within
the 95% LOA) (Figure 5-4). However, inspection of the LOA indicated that the 95%
confidence interval around the bias was generally large. The tightest 95% LOA occurred
during the step down (-5.3°, 5.6°), while the widest 95% LOA occurred during cutting
(-11.8° to 3.9°). The spread in data during cutting may have been the result of the body
rotation that naturally occurs during this task. Out of plane motion during tasks that
involve a change in direction would be expected to affect the accuracy of the 2D
measures of trunk motion.
The validity results for 2D trunk motion are consistent, in part, with previous
literature in this area.
72,131
Our findings for the triple hop (r = 0.92) (Figure 5-3) are in
general agreement with Kingston et al. who reported a moderate absolute correlation
coefficient (r = 0.65) during a similar task.
72
However, our results for the step down (r
= 0.90) (Figure 5-3) conflict with the results of Kingston et al.
72
and Schurr et al.
131

who reported no significant correlations between 2D and 3D measures of trunk motion
during single limb squatting. It should be noted however that Kingston et al.
72
reported
a weak absolute correlation coefficient (r = 0.42) with borderline significance (p =
0.087). In addition, Schurr et al.
131
examined trunk motion displacement while the
current study examined trunk position at a singular time point (i.e., peak knee flexion).
As such, caution should be taken in making direct comparison among studies.
The Pearson correlation coefficients related to the validation of 2D pelvis motion
ranged from moderate to strong, with the smallest values being observed during
deceleration and largest for the triple hop (Figure 5-1). In addition, absolute agreement
for all tasks was below 5° and there was no systematic bias (as 0 was within the 95%
LOA) (Figure 5-2). As found with the trunk however, the 95% confidence interval
around the bias was generally large. The tightest 95% LOA occurred during the triple
hop (-6.3°, 6.4°), while the widest 95% LOA occurred during cutting (-12.8° to 21.3°).

47
As noted above for the trunk, the lower correlation coefficients and/or wider 95% LOA
may be explained by trunk rotation that naturally occurs during change of direction
tasks such as cutting and deceleration.
To date, previous validation studies for kinematic measures related to pelvis
motion have only evaluated running,
37,85
so direct comparisons of the current study
findings to existing literature is limited. Our positive associations between 2D and 3D
frontal pelvis motion for all tasks evaluated coincide with the findings of Dingenen et
al. who reported that 2D and 3D pelvis drop were correlated during the stance phase
during running.
37
However, our findings conflict with Maykut et al., who reported that
2D and 3D pelvis motion during running were not correlated.
85
Maykut et al. suggest
that the differing frame rates for the 2D and 3D motion capture (60 Hz and 240 Hz,
respectively) may have been responsible for their finding of a lack of agreement.
The findings of the current study have clinical implications. First, 2D measures
of trunk and pelvis motion provide reasonable estimates of 3D motion across a wide
range of functional tasks. Importantly, the current results indicate that 2D video
methods may be appropriate for tasks that involve a change in direction. However,
when high agreement or accuracy is required with 3D angles, 2D measures of the pelvis
and trunk should be used with caution, particularly when there is body rotation.
Nonetheless, 2D video may be useful for screening of persons who may be at risk for
lower extremity injury. Future research is needed to determine if 2D measures of pelvis
and trunk motion during high demand tasks has predictive value with respect to lower
extremity injury.
There are several limitations of the current study that should be acknowledged.
First, only healthy participants were evaluated. As such, our findings cannot be
generalized to various patient populations. Second, we only assessed 2D associations
with 3D kinematic variables, using univariate analysis. It is possible that a multivariate
regression approach with the addition of other 2D measurements (such as trunk or
pelvis rotation) may have resulted in higher predictability. Third, ours was a cross-

48
sectional study, so our results do not make any assumptions of what 2D angles
constitute increased injury risk. Fourth, although all reported correlations were
statistically significant with moderate to very strong effect sizes, the clinical relevance
of our findings remain unknown and should be the focus of future investigations in this
area. Fifth, despite the fact that measurement reliability was established for our 2D
pelvis and trunk measures, the reliability of the corresponding 3D measures was not
evaluated in the current study. This could have led to diminished agreement between
the 2D and 3D measures for some tasks. Lastly, we did not evaluate pelvis or trunk
displacement. As such, our findings are only applicable to singular measurements at
peak knee flexion.

Conclusion
The results of the current study revealed that 2D frontal plane measures at the
trunk and pelvis were moderately to strongly correlated with their respective 3D angle
across a wide range of tasks. These findings suggest that 2D video analysis can be used
as an alternative to 3D motion analysis to assess frontal plane motion of the trunk and
pelvis. However, the ability of 2D trunk and pelvis angles to measure the corresponding
3D angles with high degrees of accuracy is limited, suggesting that 2D measurements
should be used cautiously when high levels of agreement or accuracy are required.

49
CHAPTER 6:
DOES THE 2D FRONTAL PLANE PROJECTION ANGLE PREDICT FRONTAL
PLANE KNEE MOMENTS DURING STEPPING, LANDING, AND CHANGE OF
DIRECTION TASKS?

In Review

Introduction
The frontal plane projection angle (FPPA) is a two-dimensional (2D) clinical
measure that was developed to identify knee valgus during dynamic tasks.
90,159
Although
the FPPA has been questioned in terms of being able to predict non-contact ACL
injury,
101,116
this measurement has been shown to distinguish between persons with and
without patellofemoral pain
52,57,158
and predict acute lower-extremity injuries.
122
Given
the potential clinical usefulness of the FPPA, there has been interest in understanding
its biomechanical utility in relation to traditional laboratory based measures of frontal
plane knee kinematics.
To date, several studies have compared FPPA measurements and 3D knee
kinematics during various tasks. Across studies, the association (R
2
) between the FPPA
and 3D knee valgus angle has been reported to range from 0% to 64% across a wide
range of tasks (i.e., single limb squat, drop jump, single leg hop, single leg land, lateral
jump, and cutting).
3,10,58,72,90,93,130,131,158
Although some authors have found that the FPPA
and 3D knee valgus are correlated, the reported agreement between these angular
measures is poor.
83
More specifically, the FPPA has been shown to overestimate true
frontal plane knee motion during a single leg squat,
131
drop jump,
72
and single leg hop,
72

with the 95% limits of agreement ranging from -30° to 17°.
72,131

The poor agreement between the FPPA and 3D frontal plane knee valgus can
be explained by previous research that has shown that what appears as knee valgus on
2D video actually is a combination of sagittal, frontal, and transverse motions at the

50
hip and knee.
1,121,158
For example, studies have reported that individuals who exhibit
poor frontal plane knee alignment based on visual assessment during a step down or
single leg squat have increased hip adduction,
121
hip flexion,
121
knee external rotation,
121

and hip internal rotation.
1
Furthermore, an increased FPPA has been found to be
correlated with increased hip adduction, knee external rotation, and hip external
rotation during a single leg squat.
158

While it is readily apparent that out-of-plane motions at the hip and knee
compromise the ability of the FPPA to accurately represent frontal plane knee
kinematics, these frontal and transverse rotations of the thigh and tibia segments may
influence variables used to calculate the frontal plane knee joint moment using inverse
dynamic equations (e.g., joint center location, joint angular velocities, segment
accelerations, etc.). To date, two studies have evaluated the relationship between the
FPPA and knee valgus moments with mixed results.
58,93
Herrington et al. (2017)
reported a strong relationship between the FPPA and peak knee valgus moment during
the single leg step down (R
2
= 42%) but not the single leg landing (R
2
= 13%).
58

Similarly, Mizner et al. (2012) reported a strong association between the FPPA and
knee valgus moment at peak knee flexion during a double-leg drop jump (R
2
= 35%).
93

To date, the ability of the FPPA to predict frontal plane knee moments during tasks
that involve pivoting and/or change of direction is not known. This is important as
such movements have been shown to result in higher knee valgus moments when
compared to tasks that are more linear in nature.
24

The purpose of the current study was to comprehensively evaluate the ability of
the FPPA to predict the frontal plane knee kinetics (peak moment, average moment,
and moment at peak knee flexion) across a wide range of tasks (stepping, landing, and
change of direction). We hypothesized that an increased 2D FPPA would be predictive
of frontal plane knee moments (i.e., increased knee valgus moments or decreased knee
varus moments). Information gained from this study will advance knowledge about the

51
clinical utility of the FPPA in characterizing movement behavior that may expose
individuals to lower-extremity injury.

Methods
Participants
Thirty-nine healthy athletes between the ages of 13 and 40 years partook in this
study (15 males and 24 females; Table 3-1).
143-145
All participants were currently
partaking in a sport with high levels of jumping, cutting, or lateral movements (such as
soccer, basketball, volleyball, lacrosse, football, netball, or tennis). Participants were
excluded if they had current lower-extremity pain, any history of ACL reconstruction,
lower-extremity injuries/surgeries in the prior 6 months, or indicated any medical
condition that would impair their ability to perform the athletic tasks.

Sample-Size Calculation
A sample size calculation was performed in G*Power (Version 3.1) based on
pilot data to determine the number of participants needed to assess the relationship
between the FPPA and frontal plane knee moment across 6 tasks. Using a 5%
significance level, 90% power, R
2
value of 0.30 (based on pilot data), and 1 predictor,
a minimum of 27 participants was deemed necessary.

Instrumentation
Three-dimensional and 2D kinematic data were collected at 120 Hz using a
video-based 8-camera motion analysis system (Simi Reality Motion Systems GmbH,
Unterschleissheim, Germany). One of the 8 cameras was positioned 80 cm off the
ground (perpendicular to the force plate) and was used to collect the required frontal
plane images for the 2D analysis.
Ground reaction forces were collected at 1200 Hz (Model #BP600900-2000,
Advanced Mechanical Technology, Inc, Watertown, MA, USA) and synchronized with

52
the motion capture system. The force plate was embedded into the floor and was used
for 5 out of the 6 tasks evaluated. For the step-down task described below, a portable
force plate was integrated into a 22 cm step (Model #O60-7000, Advanced Mechanical
Technology, Inc, Watertown, MA, USA).

Procedures
Prior to data collection, participants were informed about the nature of the study
and written consent was obtained as approved by the Institutional Review Board of the
Health Sciences Campus at the University of Southern California (Los Angeles, USA).
Once informed consent was obtained, participants warmed up on a stationary bike for
5-10 minutes. For all data procedures outlined below, data were obtained on the right
limb.
Participants were instrumented with 17 reflective markers (10 mm diameter) on
the right lower extremity, as previously described.
143,145
Two-dimensional video and 3D
motion analysis were collected during the following tasks: 1) Step Down, 2) Drop
Jump, 3) Lateral Shuffle, 4) Deceleration, 5) Triple Hop, and 6) Side-Step-Cut.
Details regarding the instructions provided to participants for each of the tasks can
found in Table 3-1.
143-145
These tasks were selected based on current knowledge of
movements thought to be associated with various sport injuries. A trial was considered
successful if all markers remained visible and only the foot of tested limb fully contacted
the force plate. Participants were permitted to practice until comfortable with the
performance of each task. One to 2 trials were obtained for each of the tasks above.

Data Analysis
The first successful trial was selected for each task and used for data analysis.
Marker position data were labeled in Simi Motion and then exported with the force
data to Visual3D software (C-Motion, Inc, Germantown, MD, USA). Marker
trajectory and analog force plate data were low-pass filtered at 12 Hz, using a fourth-

53
order Butterworth filter.
150
Joint angles were calculated using a X-Y-Z (sagittal-frontal-
transverse) Cardan sequence.
Inverse dynamics equations were used to calculate net joint moments (external)
at the knee. Moment data were normalized to body mass and height. Three frontal
plane knee moment variables were extracted (peak moment, average moment, and
moment at peak knee flexion). The peak and average frontal plane knee moments were
calculated during the deceleration phase of all tasks (initial contact to peak knee
flexion). In addition, the frontal plane knee moment at peak knee flexion was
identified. For the step down, the peak and average frontal plane knee moments were
calculated during the lowering phase (initiation of the movement to the time at which
the heel touched the ground). For calculation of the peak moment for trials in which
a valgus moment was not present, the minimum varus moment was identified and used
for statistical analysis.
For the 2D video analysis, the image containing peak knee flexion was identified.
For the step down, the image at which the contralateral heel touched the ground was
used for analysis. Images were uploaded into ImageJ software (Version 1.50i, National
Institute of Health, USA) for 2D angle assessments. The FPPA was measured as the
angle formed by 3 points (ASIS, knee joint center, ankle joint center). This value was
subtracted from 180 to represent the anatomical frontal plane alignment of the knee.
72

A positive value represented knee valgus (knee joint center medial to a line formed
from the ankle and ASIS) and a negative represented knee varus (knee joint marker
lateral to a line formed from the ankle and ASIS) (Figure 2-5). All 2D measurements
were obtained by a single investigator who demonstrated excellent intra-rater reliability
for all tasks prior to the start of the study (ICCs ranging from 0.91 to 1.0).

Statistical Analysis
Linear regression analysis was used to assess the ability of the 2D FPPA angle
(independent variable) to predict the frontal plane knee moment (dependent variable).

54
This analysis was repeated for each task and was run separately for each dependent
variable (peak frontal plane knee moment, average frontal plane knee moment, and
frontal plane knee moment at peak knee flexion). R
2
values were interpreted as strong
(>= 0.50), moderate (0.25-0.49), weak (0.10-0.24), and negligible (0.0-0.09).
95
All
statistical analyses were performed using SPSS Version 27 (Chicago, Illinois, USA) and
a custom MATLAB script (The Mathworks, Inc., Natick, MA) with alpha set at 0.05.

Results
Due to technical issues with the force plate, ground reaction force data were not
available for one subject during the drop jump and 8 participants during the step-down
task. Descriptive statistics for the FPPA, peak frontal plane knee moment, and average
frontal plane knee moment for each task are presented in Figure 6-1. Time series data
for the frontal plane knee moment are presented in Figure 6-2.

Figure 6-1. Average FPPA and moment variables for the 6 tasks evaluated. Error bars represent 1 SD.

55

Relationship between FPPA and Peak Frontal Plane Knee Moment
The FPPA was found to significantly predict the peak frontal plane knee moment
for deceleration (R
2
= 0.12, p = 0.032) and side-step-cut (R
2
= 0.25, p = 0.001), with
a larger FPPA predicting increased knee valgus moments (or decreased knee varus
moments). However, the FPPA did not predict the peak frontal plane knee moment
for step down, drop jump, lateral shuffle, and triple hop (Figure 6-3).

Figure 6-2. Time-normalized frontal plane knee moment data for the 6 tasks evaluated. Error bars
represent 1 SD. Positive values represent knee valgus moments.

56

Relationship between FPPA and Average Frontal Plane Knee Moment
The FPPA was found to significantly predict the average frontal plane knee
moment for drop jump (R
2
= 0.25, p = 0.001), shuffle (R
2
= 0.40, p < 0.001),
deceleration (R
2
= 0.20, p = 0.004), triple hop (R
2
= 0.15, p = 0.015), and side-step-
cut (R
2
= 0.31, p < 0.001), with a larger FPPA predicting increased knee valgus
moments (or decreased knee varus moments). However, the FPPA did not predict the
average frontal plane knee moment for step down (R
2
= 0.0, p = 0.775) (Figure 6-4).
Figure 6-3. Linear regression models to predict the peak frontal plane knee moment for each task.

57

Relationship between FPPA and Frontal Plane Knee Moment at Peak Knee Flexion
The FPPA was found to significantly predict the frontal plane knee moment at
peak knee flexion for drop jump (R
2
= 0.39, p < 0.001), shuffle (R
2
= 0.45, p < 0.001),
deceleration (R
2
= 0.16, p = 0.013), triple hop (R
2
= 0.17, p = 0.008), and side-step-
cut (R
2
= 0.27, p < 0.001), with a larger FPPA predicting increased knee valgus
moments (or decreased knee varus moments). However, the FPPA did not predict the
frontal plane knee moment at peak knee flexion for step down (R
2
= 0.02, p = 0.41)
(Figure 6-5).
Figure 6-4. Linear regression models to predict the average frontal plane knee moment for
each task.

58

Discussion
The purpose of the current study was to evaluate the ability of the 2D FPPA to
predict the frontal plane knee moment (average, peak, and moment at peak knee
flexion) during stepping, landing, and change of direction tasks. In general, the FPPA
was a better predictor of the average frontal plane knee moment (5 out of 6 tasks) and
frontal plane knee moment at peak knee flexion (5 of 6 tasks) compared to the peak
frontal plane knee moment (2 out of 6 tasks). For all significant models, an increased
FPPA predicted increased knee valgus moments (or decreased knee varus moments)
during landing and change of direction tasks (but not stepping). However, the strength
Figure 6-5. Linear regression models to predict the frontal plane knee moment at maximum
knee flexion for each task.

59
of the predictive models was weak to moderate (R
2
= 12% to 45%), highlighting that
the utility of the FPPA as an indicator of frontal plane knee moments during landing
and change of direction tasks is limited.
Our results are in general agreement with the findings of Herrington et al.
(2017)
58
and Mizner et al. (2012),
93
both of which examined the ability of the FPPA to
predict knee valgus moments during various tasks. Mizner et al. reported that an
increased FPPA predicted the knee valgus moment at peak knee flexion during a drop
jump task (R
2
= 35%),
93
which is comparable to our moment results at peak knee
flexion for the drop jump (R
2
= 39%). Herrington et al. reported that an increased
FPPA did not predict the peak knee valgus moment during a single leg landing from a
box (R
2
= 13%),
58
which agrees with our finding for the peak frontal plane knee moment
during the triple hop (R
2
= 2%). However, Herrington et al. reported that an increased
FPPA predicted the peak knee valgus moment during a single leg squat (R
2
= 42%),
58

which is in contrast with our findings for the step down for the peak frontal plane knee
moment (R
2
= 1%). However, the step down and single leg squat differ in a number of
kinematic variables,
81
which makes direct comparisons difficult.
Across tasks, the highest R
2
values were found for the average frontal knee
moments and frontal plane knee moments at maximum knee flexion. Given that the
FPPA was measured at peak knee flexion, it is logical that the FPPA was predictive of
the frontal plane knee moment at that point in time. Additionally, the fact that peak
knee flexion was used to indicate the end of the deceleration phase for each task may
explain why the FPPA predictive models for the average moment during the
deceleration phase were similar to those observed for the frontal plane knee moment
at peak knee flexion. The ability of the FPPA to predict the peak frontal plane knee
moment was limited to 2 of the 6 tasks (deceleration and cutting), with R
2
values being
lower than the other 2 variables examined. The limited ability of the FPPA to predict
the peak frontal plane knee moments may be explained by the fact that the peak
moment did not always occur at the same time point at which the FPPA was measured

60
(Figure 6-2). As such, the timing of the kinetic variables of interest should be considered
when measuring the FPPA at a single point in time.
With respect to the strength of the predictions across tasks, the step down
exhibited non-significant results for all 3 frontal plane knee moment variables (R
2
= 0-
2%) (Figures 6-3, 6-4, 6-5). This finding may be related to the fact that 100% of
participants exhibited average knee varus moments during this movement, and this task
had the lowest average frontal plane knee moment (Figure 6-1, Figure 6-4). In contrast,
the strongest significant relationship was observed for the shuffle task, which had the
second highest average frontal plane knee moment and a relatively large prevalence of
average knee valgus moments (69% of participants) (Figure 6-1, Figure 6-4). It appears
that the FPPA may be a stronger predictor of frontal plane knee kinetics when a knee
valgus moment is present, with the strength of the predictability contingent on the
observed frequency and magnitude of knee valgus moments. This is logical as the FPPA
is indicative of inward collapse of the knee and therefore would be expected to be
indicative of the variables that would be related to a knee valgus moment (i.e., medial
positioning of the knee joint center, etc.).
Previous studies have reported that the FPPA is an inconsistent predictor of
frontal plane knee kinematics
3,10,58,72,90,93,130,131,158
and that the general agreement between
2D and 3D frontal plane knee angles is poor.
83
Based on the current study and the
work of previous authors who have evaluated the ability of the FPPA to predict frontal
plane knee moments,
58,93
it appears that the FPPA may be a better indicator of knee
kinetics as opposed to knee kinematics. It is possible that the clinical utility of the
FPPA as a predictor of injury
122
or the ability of the FPPA to differentiate between
healthy and clinical populations
52,57,158
may lie in the fact that this measure is a predictor
of frontal plane knee moments. An argument could be made that the frontal plane knee
moment is more suggestive of knee loading as opposed to frontal plane knee motion.
There are several limitations within the current study that warrant discussion.
First, our data were obtained from healthy individuals. As such, our results may not be

61
applicable to those with specific knee conditions (i.e., patellofemoral pain, ACL injury,
etc.). Second, only the deceleration or lowering phase of each task was considered in
our moment analysis. Therefore, our results may not apply to the acceleration phase
of the tasks evaluated. Third, the current study was cross-sectional in nature. Our results
cannot be interpreted to suggest that increased FPPA angles are predictive of knee
injury. Lastly, for all regression models, only a single predictor (FPPA) was examined.
The R
2
values reported could perhaps be improved by including other 2D measurements
such as frontal plane motion at the hip, pelvis, or trunk.
155

Conclusion
The results of the current study suggest that the FPPA is a predictor of frontal
plane knee loading during landing and change in direction tasks, specifically when the
frontal plane knee moment is calculated as the average moment or the moment at peak
knee flexion. For all significant models, an increased FPPA (indicative of medial knee
collapse) predicted increased knee valgus moments (or decreased knee varus moments)
during landing and change of direction tasks (but not stepping). However, the ability
of the FPPA to predict frontal plane knee kinetics appears to be task dependent, with
the strength of the prediction improved with increased frequency and magnitude of
observed knee valgus moments. In addition, the strength of the prediction was weak to
moderate, highlighting that the validity of the FPPA as a predictor of frontal plane
knee moments during landing and change of direction tasks is limited.

62
CHAPTER 7:
PREDICTION OF ACL RE-INJURY IN FEMALES USING 2D VIDEO:
A RETROSPECTIVE CASE-CONTROL STUDY

Introduction
Approximately 1 in 4 athletes who return to a high-risk sport after primary ACL
reconstruction (ACLR) will go on to sustain another ACL injury.
109,156
Specifically,
athletes who return to sport have a 30 to 40 times greater risk of sustaining a second
ACL injury compared to their uninjured counterparts.
156
Prospective studies have
identified specific movement impairments at the trunk, pelvis, hip, and knee that are
predictive of initial ACL injury.
59,77,79,80
Not surprisingly, movement impairments at the
trunk, pelvis, hip, and knee also have been shown to be predictive of a second ACL
injury.
70,71,110,119

The fact that a wide range of movement impairments have been identified as
being contributory to primary and secondary ACL injury,
59,70,71,77,79,80,110,119
highlights the
need to consider evaluation of movement quality as part of the return to sport decision
process for athletes who have undergone ACLR. One clinic-based movement
assessment used to assess ACL injury risk is the Landing Error Scoring System
(LESS).
105
Although the LESS has been shown to predict initial ACL injury in youth
soccer players, predictive capacity has not been replicated in high school and collegiate
athletes.
104,140
A notable shortcoming of LESS is the fact that only a single task is
evaluated (double limb landing task). This limitation is important as ACL injuries
typically occur during single-limb, multidirectional tasks.
4,29

The lack of a comprehensive movement assessment for patients who have
experienced an ACL injury led to the development of the video-based Movement
Performance Assessment (MPA). The MPA examines movement features during a
range of sport-specific tasks typically associated with ACL injury. Movement quality is
quantified in terms of 5 movement domains that have been hypothesized to contribute

63
to ACL injury (knee strategy, shock absorption, knee stability, pelvis stability, and trunk
stability). Although the video-based measurements that are used to quantify movement
impairments for each movement domain have been reported to be predictive of
corresponding 3D kinematic and/or kinetic variables,(Chapter 3-6)
143-145
it remains
unknown which, if any, of the MPA movement constructs are predictive of ACL re-
injury.
Given the need to improve return to sport outcomes following ACLR, the
purpose of this retrospective case-control study was two-fold: 1) to determine if a
subgroup (or cluster) of female athletes who exhibit a consistent high-risk movement
profile across tasks can be identified within each of the MPA movement domains, and
2) to determine if athletes assigned to the high-risk subgroup within each of the MPA
movement domains are at greater risk of ACL re-injury compared to those assigned to
lower risk subgroups. Data collected as part of this study is the first step in the
development of a valid clinic-friendly assessment to aid in the identification of persons
at risk for re-injury following ACLR.

Methods
Participants
The patients in this retrospective case-control study were female athletes post
ACLR who had previously completed the MPA as part of return to sport testing for
the surgical limb between March 2011 through December 2019. Patients were identified
from a clinical database and were sent a survey if the following criteria were met: (1)
at least 12 months had passed since return to sport testing or (2) we had a record of
the patient sustaining a subsequent ACL injury after baseline return to sport testing.
Based on these criteria, 345 patients were identified and sent an injury survey (see
below for details).

64
Sample-Size Calculation
The minimum sample size was estimated based on estimates for cluster analysis
and logistic regression. For clustering (or separating the sample into subgroups), the
minimal sample should be no less than 2
k
(k = number of variables).
45
Each clustering
analysis required a maximum of 6 variables, suggesting a minimum sample of 64
participants. For logistic regression with a binary risk factor (i.e., ability of cluster
subgroup to predict ACL re-injury), a sample size calculation based on the score test
was performed using 6 input variables.
47,48
The prevalence of high risk movement in
controls was considered 50%, with an odds ratio (OR) of 4.0 (within 2.1 to 8.4 range
for biomechanical risk factors for ACL re-injury).
110,119
Assuming 25% risk of re-
injury,
109,156
1:2 matching ratio of case to controls, 80% power, and 5% significance
level, a minimum of 70 subjects (23 cases and 47 controls) were required.

Procedures
Eligible patients were contracted via email, phone and/or text messaging to
participate in an online survey concerning their ACL injury status and athletic
participation after return to sport (Appendix). Electronically signed informed consent
was obtained prior to survey commencement, as approved by the Institutional Review
Board of the Health Sciences Campus at the University of Southern California. Only
survey results from those patients who had returned to their sport within 6 months of
testing were used for analysis. For those who reported an ACL re-injury (ipsilateral or
contralateral), only non-contact injuries that occurred within 36 after returning to sport
were considered. Each injured case was matched with 2-3 non-injured controls
148
based
on graft type, age (within 5 years), highest level of sport returned (or injured), and
athletic exposures (Table 7-1).

65
Table 7-1. Matching Criteria for Cases & Controls*

Of the 345 female patients contacted, 184 started the survey (53% response rate)
(Figure 7-1). After elimination of patients without clear return to sport status, 159
remained (48 injured and 111 non-injured). From the injured cohort (n=48), 40 non-
contact injured patients were identified (25% re-injury rate after returning to sport via
non-contact mechanism), resulting in 23 cases. From the non-injured cohort (n=111),
61 controls were identified. Descriptive data for the final female cohort (23 cases and
61 controls) are shown in Table 7-2.

Item

Criteria

Example

Age

Match age within 5 years.

12-year case matched with 7-to-17-year
control.

Sport
Level
1

Match highest level of sport returned or
injured.
Soccer case (level 1) matched with
basketball (level 1) control.

Graft Type Match graft type (allograft vs autograft).

Allograft case matched with allograft
control.

Athletic
exposures
2

Match yes/no responses to participation
in at least 50 hrs of pivoting and/or
cutting sport during each year period
after RTS. Control allowed to have
higher participation than case (but not
vice versa). Time frame for matching
based on months from RTS to injury for
case. Months were rounded down (e.g., 6
months rounded to 0 years) or up (e.g., 7
months rounded to 1 year).

Case injured 7 months after RTS with
participation in at least 50 hours of
pivoting/cutting during year 1 (i.e.,
“yes”), matched with control with same
level of participation during year 1 (i.e.,
“yes”).

*RTS, return to sport.
1
Level 1 is a sport with frequent pivoting, cutting, or jumping (e.g., football, soccer, volleyball,
lacrosse). Level 2 is a sport with less pivoting, cutting, or jumping (e.g., baseball, skiing, ice hockey,
tennis).
2
Survey question: “Once your returned to your sport or activity, did you participate in a pivoting
and/or cutting sport for at least 50 hours during the following times? Year 1, Year 2, Year 3, etc.”

66

Figure 7-1. Flow chart of 345 female athletes contacted to complete survey, which resulted in 23 cases and 61 controls. RTS: return to
sport.

67
Table 7-2. Patient Demographics (N = 84)*

Data Analysis
2D Angular measures were scored at peak knee flexion for all matched
cases/controls according to the procedures outlined in the validation studies described
in Chapters 3-6 (Figure 2-1 to 2-5).
143-145
In total, 27 angular measures were obtained for
analysis across the 6 tasks conceptualized for the MPA and were subsequently divided
among 5 separate movement domains (Figure 7-2). Angular measures were scored by 2
raters who demonstrated acceptable intra-rater and inter-rater reliability, as determined
by the intraclass correlation coefficient (ICC). Specifically, ICCs for intra-rater
reliability ranged from 0.74-1.0 and 0.63-0.99 for inter-rater reliability.

Cases
(n = 23)
Controls
(n = 61)

P Value
3

Age at RTS testing (yrs)

16.5 (2.2), 14-22 16.7 (2.6), 11-26 0.77

BMI (kg/m
2
) 22.1 (2.4), 18.8-29.3 23.3 (2.9), 19.3-37.0 0.07

Graft Type
Autograft
Allograft

17 (74)
6 (26)

46 (77)
15 (25)

0.89
Sport, Level 1
1

23 (100%) 61 (100%) --

Athletic exposures (mo)
2

9.5 (10.9), 0-27 9.6 (10.8), 0-27 0.96
Time surgery to RTS (mo)

8.8 (1.7), 6-12

10.4 (3.5), 5-24

0.007
Time RTS testing to RTS (mo)

1.3 (1.5), 0-5

1.4 (1.8), 0-6

0.78
Time RTS to injury (mo)

9.5 (10.9), 0-27 --
2+ Prior ACL Tears
Yes
No

1 (4)
22 (96)

11 (18)
50 (82)

0.17

Injured Side
Ipsilateral
Contralateral

9 (39)
14 (61)

--
1
Level 1 is a sport with frequent pivoting, cutting or jumping (e.g., football, soccer, volleyball,
lacrosse).
2
Athletic exposures defined in Table 7-1. Months from RTS to injury for cases used as athletic
exposures for control.
3
Independent T-test for continuous measures and Pearson’s chi square for categorical. Fisher’s exact
test was used instead of the chi-square test if the number of patients in a cell was below 5.
*RTS, return to sport. For subjects with a missing RTS month, January was assumed. If the RTS
evaluation was later than this date, the evaluation date was used. If the RTS date was completely
missing, the evaluation date was used. Values are mean (SD), range or group no. (%).

68

Statistical Analysis
To determine if a high-risk subgroup (or cluster) of athletes who exhibit a
consistent movement profile across tasks could be identified within each of the MPA
movement domains, cluster analysis was used (study aim 1). Cluster analysis was
conducted using the k-means method (based on the squared Euclidean distance
measure) to partition participants into mutually exclusive groups (or clusters) based on
a combination of measures.
62,82,161
This procedure was conducted separately for each
movement domain using 5-6 variables (tasks) per analysis (Figure 7-3). Since clustering
is sensitive to data order, the best arrangement (i.e., solution with the lowest total sum
of point-to-centroid distances) out of 100 random initializations was selected as the
final solution. Prior to clustering, all variables were standardized to z scores with a
mean value of 0 and a SD of 1 to equalize the importance of each variable. Data were
checked for extreme outliers (>5 SDs from the mean). The “optimal” number of
clusters for each movement domain was determined by comparing 2 to 5 cluster
solutions using the silhouette criterion to assess the quality of cluster separation. The
silhouette coefficient ranges from -1.0 to 1.0 and was interpretated as good (0.50 to
1.0), fair (0.20-0.49), or poor (-1.0 to 0.19).
66

The integrity of the clustering results for each movement domain was visualized
using scatter plots and principial component analysis (PCA) to reduce multi-
dimensional data (i.e., 5 to 6 dimensions) to lower dimensions (i.e., final cluster
number). Within each movement domain, differences between subgroups in the 2D
Figure 7-2. The MPA is comprised of 5 separate movement domains, each which contains 2D metrics
for 5-6 tasks. In total, 33 angular measures are required, which results in 27 measures for data analysis.

69
angular metrics were assessed using one-way ANOVAs or independent t-tests
(depending on the number of clusters identified) to inform cluster interpretation. Based
on previous literature related to ACL injury risk (prospective as well as cross-sectional
studies), the “high risk” subgroups were operationally defined as follows: high knee
valgus (defined by the highest average frontal plane projection angle (FPPA) across
tasks within the knee stability domain), high pelvis drop (defined by the highest average
pelvis drop across tasks within the pelvis stability domain), high ipsilateral trunk lean
(defined by the highest average ipsilateral trunk lean across tasks within the trunk
stability domain), high knee extensor bias (defined by the lowest average trunk-tibia
angle across tasks within the knee strategy domain), and high impact forces (defined
by the lowest average thigh angle across tasks within the shock absorption domain).
Unconditional logistic regression was used to determine if the subgroup
operationally defined as “high risk” within each movement domain was predictive of
ACL re-injury compared to lower risk subgroups (study aim 2).
111
This analysis was
performed separately for each movement domain. Prior to the main analysis, the
influence of 6 non-movement variables on injury status was assessed using logistic
regression (i.e., time surgery to return to sport, athletic exposures, age, graft type, body
mass index (BMI), 2+ prior ACL tears). Any non-movement factor that individually
predicted injury status at p<0.15 was deemed a co-variate and controlled for in the
main analysis. This procedure identified 3 covariates: time surgery to return to sport,
BMI, and 2+ prior ACL tears.
To determine if inclusion of a significant movement domain enhanced
prediction performance beyond a baseline model containing 3 covariates, several
measures were computed.
21
The Chi-square Likelihood ratio test was used to compare
the model fit for two logistic models. Since the Likelihood ratio test is considered the
most powerful statistical test for determining if there is any improvement in prediction
performance with the inclusion of a new predictor,
21,54,115
this was the primary statistical
test of interest. The extent of model improvement was quantified with two additional

70
measures: change in area under the curve (AUC) and integrated discrimination
improvement (IDI) index. The AUC for a receiver operating characteristic curve
(ROC) for the model with and without the movement domain variable were compared.
ROC curves were constructed using predicted probabilities from the logistic models.
The improvement in AUCs was assessed by DeLong’s test for two correlated
ROC curves (one-tailed test).
33
The AUC was interpretated as excellent (0.90-1.0),
good (0.80-0.89), fair (0.70-0.79), poor (0.60-0.69), and fail (0.50-0.59). The
contribution of the movement domain to risk prediction beyond the baseline model
was further assessed by the IDI index (one-tailed test).
21,112-114
The IDI summarizes the
average change in probabilities among cases and controls. A useful novel predictor
results in increased risk for cases and decreased risk for controls, with larger IDI values
equating to improved predictive value. All analyses were performed using SPSS
statistical software (SPSS Inc., Chicago, Ill; Version 27), MATLAB (The Mathworks,
Inc., Natick, MA; Version 2021a), and R (The R Project for Statistical Computing,
Vienna, Austria; Version 4.1.2) with alpha set at 0.05. P values <= 0.05 were considered
significant.

Results
Clustering
No outliers were found based on the criterion above. The 2-cluster solution
maximized the silhouette criterion for all movement domains and was thus considered
the “optimal” solution. The quality of separation for the 2-cluster solution (i.e.,
silhouette value) was fair for all movement domains, ranging from 0.33 (knee stability)
to 0.47 (shock absorption) (Figure 7-4). Therefore, k-means clustering was conducted
to separate athletes in each movement domain into 2 subgroups (k=2).

71

The clustering and PCA results for each movement domain are presented in
Figure 7-4. Within each movement domain, the underlying 2D angular measurements
for Cluster 1 vs Cluster 2 differed significantly from each other as determined by
independent t-tests (all p-values <= 0.03) (Figure 7-5, Table 7-3). Given that there was
a consistent pattern across angles for each movement domain (e.g., cluster 1 contained
higher angles on average than cluster 2, or vice versa), cluster characterization was
possible. As such, 2 subgroups were inferred for each movement domain: (1) high vs.
low knee extensor bias (within knee strategy domain), high vs. low impact forces (within
shock absorption domain), (3) high vs. low knee valgus (within knee stability domain),
high vs. low ipsilateral trunk lean (within trunk stability domain), and (5) high vs. low
Figure 7-3. Optimal number of clusters for each movement domain was determined by finding
the cluster solution with the maximum silhouette value. The silhouette value was maximized
with a 2-cluster solution (k=2) for all movement domains.

72
pelvis drop (within pelvis stability domain) (Table 7-3). The distribution of cases and
controls within each subgroup is presented in Figure 7-4.

Figure 7-4. Visualization of k-means clustering output (k=2) after using principial component analysis
to reduce multi-dimensional data to two-dimensions. K-means clustering was performed separately for
each MPA movement domain, with 84 athletes.

73

Figure 7-5. Cluster characteristics in female athletes post ACLR. K-means clustering (k=2) was
performed separately for each MPA movement domain with 5-6 variables (tasks). Each row (color)
represents a movement domain, and each column represents a task. Athletes are consistent across rows
only. C1: Cluster 1; C2: Cluster 2. The characterization for each cluster (which differs based on
movement domain) is shown in Table 7-3. Knee strategy is scored using the trunk-tibia angle (+trunk
more forward than tibia, -tibia more forward than trunk). Shock absorption is scored using the thigh
angle (+increased hip and/or knee flexion). Knee stability is scored using the frontal plane projection
angle (+valgus, -varus). Trunk stability is scored using trunk lean (+ipsil, -contra). Pelvis stability is
scored using pelvis tilt (+drop, -ipsil).

74

Table 7-3. 2D Measurements (degrees) within Clusters for Each Movement Domain (N = 84)

Movement
Domain*

Variable

Cluster 1

Cluster 2

T-test
Knee Strategy
(Trunk-Tibia,
deg)
Cluster
Characterization
High knee extensor bias
(n=46)**
Low knee extensor
bias (n=38)

Step down
Drop jump
Shuffle
Deceleration
Triple hop
Side-step-cut

-2.1 (14.3)
15.4 (10.4)
18.4 (11.7)
28.9 (14.8)
17.0 (9.3)
21.1 (13.3)

13.7 (13.3)
28.3 (9.9)
32.6 (10.3)
50.9 (13.6)
35.7 (10.2)
38.2 (12.7)

p < 0.001
p < 0.001
p < 0.001
p < 0.001
p < 0.001
p < 0.001

Shock
Absorption
(Thigh, deg)
Cluster
Characterization
Low impact forces
(n=49)
High impact forces
(n=35)**

Drop jump
Shuffle
Deceleration
Triple hop
Side-step-cut

83.0 (10.1)
54.1 (7.0)
77.0 (6.6)
63.9 (8.4)
61.2 (7.8)

68.0 (8.1)
42.1 (7.6)
67.1 (5.9)
49.3 (8.3)
53.5 (9.6)

p < 0.001
p < 0.001
p < 0.001
p < 0.001
p < 0.001

Knee Stability
(FPPA, deg)
Cluster
Characterization
High knee valgus
(n=47)**
Low knee valgus
(n=37)

Step down
Drop jump
Shuffle
Deceleration
Triple hop
Side-step-cut

3.9 (9.6)
-17.6 (28.7)
-1.8 (15.0)
21.7 (14.0)
10.4 (9.9)
19.5 (19.1)

-8.6 (8.7)
-34.2 (23.9)
-18.2 (12.7)
8.8 (13.8)
1.9 (9.4)
-8.7 (18.4)

p < 0.001
p = 0.006
p < 0.001
p < 0.001
p < 0.001
p < 0.001

Trunk Stability
(Trunk lean,
deg)
Cluster
Characterization
Low ipsilateral trunk
lean (n=46)
High ipsilateral trunk
lean (n=38)**

Step down
Shuffle
Deceleration
Triple hop
Side-step-cut

2.3 (3.4)
-4.5 (8.2)
0.1 (3.9)
4.1 (6.8)
-6.4 (10.6)

10.1 (5.1)
3.1 (7.7)
6.4 (4.9)
8.7 (7.5)
1.5 (10.0)

p < 0.001
p < 0.001
p < 0.001
p = 0.003
p < 0.001
Pelvis Stability
(Pelvis tilt, deg)
Cluster
Characterization
Low pelvis drop (n=49) High pelvis drop
(n=35)**

Step down
Shuffle
Deceleration
Triple hop
Side-step-cut

2.9 (2.4)
0.9 (3.9)
-0.8 (3.2)
-2.0 (4.7)
4.7 (4.6)

7.1 (2.7)
2.9 (4.7)
3.6 (3.7)
1.6 (4.5)
12.6 (6.8)

p < 0.001
p = 0.03
p < 0.001
p < 0.001
p < 0.001

*Abbreviations. FPPA: frontal plane projection angle. Data presented as mean (SD). Knee strategy is
scored using the trunk-tibia angle (+trunk more forward than tibia, -tibia more forward than trunk).
Shock absorption is scored using the thigh angle (+increased hip and/or knee flexion). Knee stability
is scored using the frontal plane projection angle (+valgus, -varus). Trunk stability is scored using
trunk lean (+ipsil, -contra). Pelvis stability is scored using pelvis tilt (+drop, -ipsil). Values expressed
as mean (SD) for continuous metrics and no (%) for categorical metrics.
**Subgroup operationally defined as “high risk.”

75
Movement Domain Prediction
Of the 5 MPA movement domains evaluated, only knee strategy was predictive
of non-contact ACL re-injury (ipsilateral or contralateral) (Table 7-4). Within the knee
strategy domain, female athletes who favored use of the knee extensors relative to the
hip extensors (defined by a lower 2D trunk-tibia angle across tasks) were at greater risk
of ACL re-injury (Table 7-4, Figure 7-6). Compared to Cluster 2 (low knee extensor
bias), the odds of ACL re-injury were significantly increased in Cluster 1 (high knee
extensor bias) (adjusted OR = 3.19, 95% CI: 1.02, 9.96, p = 0.05).

Table 7-4. Separate Logistic Regression Models for the 5 MPA Movement Domains to Predict ACL
Re-injury in Female Athletes

Movement
Domain
Predictor Adjusted Odds Ratio [95% CI] P value

Knee
Strategy
Time surgery return to sport (mo)

0.76 [0.60, 0.97] 0.03
BMI (kg/m
2
)

0.80 [0.63, 1.01] 0.06

2+ prior ACL tears

0.28 [0.03, 2.51] 0.26
Cluster 1 (high knee extensor bias) 3.19 [1.02, 9.96] 0.05
Shock
Absorption

Time surgery return to sport (mo)

0.75 [0.59, 0.96]

0.02
BMI (kg/m
2
)

0.80 [0.63, 1.02] 0.07
2+ prior ACL tears

0.37 [0.04, 3.45] 0.38
Cluster 2 (high impact forces) 2.76 [0.92, 8.30] 0.07
Knee
Stability

Time surgery return to sport (mo)

0.78 [0.62, 0.98]

0.04
BMI (kg/m
2
)

0.78 [0.62, 0.99] 0.05
2+ prior ACL tears

0.26 [0.03, 2.32] 0.23
Cluster 1 (high knee valgus) 1.08 [0.38, 3.09] 0.88
Trunk
Stability

Time surgery return to sport (mo)

0.77 [0.61, 0.98]

0.03
BMI (kg/m
2
)

0.78 [0.61, 0.99] 0.04
2+ prior ACL tears

0.26 [0.03, 2.38] 0.24
Cluster 2 (high ipsilateral trunk lean) 1.32 [0.46, 3.76] 0.61
Pelvis
Stability

Time surgery return to sport (mo)

0.78 [0.62, 0.98]

0.04
BMI (kg/m
2
)

0.79 [0.62, 1.00] 0.05
2+ prior ACL tears

0.26 [0.03, 2.36] 0.23
Cluster 2 (high pelvis drop) 1.29 [0.46, 3.63] 0.63

76

Measures of model performance are summarized in Table 7-5. The change in the
Likelihood ratio Chi-square comparing the baseline model with 3 covariates to the
enhanced model (i.e., baseline model + knee strategy) was statistically significant,
indicating that addition of knee strategy to the baseline model improved the goodness
to fit (DLikelihood Chi-square = 4.32, p = 0.04). ROC curve analysis of the enhanced
model showed an AUC of 0.78 [95% CI: 0.67, 0.89], compared to an AUC of 0.76
[95% CI: 0.64, 0.87] for the baseline model. The 0.02 [95% CI: -0.05, 0.10] increase
in the AUC with the addition of the knee strategy domain was not statistically
significant (Z = 0.68, p = 0.25, one-tailed). However, the addition of the knee strategy
domain resulted in a statistically significant improvement in overall predictive ability,
based on an IDI of 0.06 [95% CI: 0.01, 0.11, Z = 2.27, p = 0.01, one-tailed]. The
average increased risk for cases was 4% while the average decreased risk for controls
was 2% (IDI = 4% + 2% = 6%). The change in predicted probability of ACL re-injury
for cases ranged from -17% to 12% when the knee strategy domain was added to the
baseline model (Figure 7-7). When in the presence of a high knee extensor bias, the
shift in ACL re-injury risk for cases always was positive (5% to 12%).

Figure 7-6. Within the knee strategy domain, the high knee extensor bias subgroup (defined by a lower
2D trunk-tibia angle across tasks) was at increased risk for ACL re-injury compared to the low knee
extensor bias subgroup.

77

Table 7-5. Statistics for Model Improvement with Addition of Knee Strategy Domain to Baseline
Model*

Metric Baseline
Model
P value Baseline Model +
Knee Strategy
P value
D with Knee
Strategy
Addition

P value
Likelihood
ratio chi-square

12.30 0.006 16.61 0.002 4.32 0.04
AUC

0.76
(0.64, 0.87)

<0.001 0.78
(0.67, 0.89)
<0.001 0.02
(-0.05, 0.10)
0.25
a

Overall IDI

-- -- -- -- 0.06
(0.01, 0.11)

0.01
a

IDIcases

-- -- -- -- 0.04 --
IDIcontrols

-- -- -- -- 0.02 --
*Values include 95% CIs when relevant. AUC: area under the curve; IDI: integrated discrimination
improvement. The baseline model contained 3 covariates, as highlighted in Table 7-4.
a
P-values are one-tailed for improvement in AUC and IDI.

Figure 7-7. Risk of probability for ACL re-injury in cases increased 4% on
average when the knee strategy domain was added to the baseline model
with 3 adjustment factors. The change in risk ranged from -17% to 12%.
The shift in ACL re-injury risk was always positive (5% to 12%) when in
the presence of a high knee extensor bias strategy (n=17).

78
Discussion
A wide range of movement impairments have been shown to be predictive of
primary or secondary ACL injury.
59,70,71,77,79,80,110,119
As such, there is a need to evaluate
potential movement related risk factors as part of injury risk screening and return to
sport testing. The MPA was designed as a clinic-friendly means to assess movement
impairments consistent with ACL injury. The primary purpose of this study was to
determine which of the movement domains within the MPA are predictive of ACL re-
injury (ipsilateral or contralateral) in female athletes. To accomplish this aim, we first
took a clustering analysis approach to identify high risk subgroups (i.e., clusters) within
each of the MPA movement domains. We next sought to determine if athletes assigned
to the high-risk subgroups within each of the MPA movement domains are at greater
risk of ACL re-injury compared to those assigned to lower risk subgroups.
Cluster analysis, as performed in the current study, could separate female
athletes into groups with distinct movement features. Specifically, 2 subgroups for each
movement domain were found to maximized heterogeneity (Figure 7-3). Interestingly,
movement tendencies were consistent across tasks within each movement domain when
using a 2-cluster solution, allowing the ability to distinguish between female athletes
exhibiting high vs. low-risk movement attributes (Figure 7-5, Table 7-3). For the knee
strategy domain, females in cluster 1 had lower average 2D trunk-tibia angles for all 6
tasks compared to females in cluster 2. In addition, all 2D measures of knee strategy
were statistically significant, highlighting that all tasks contributed to the clustering
outcome. Similar findings were observed for the other movement domains in that the
2D measures of interest were consistently lower in one cluster compared to the other.
Despite statistical significance however, it should be noted that the quality of separation
between clusters was fair for all domains (with the lowest quality occurring in knee
stability and the highest quality in shock absorption) (Figure 7-3). Therefore, clustering
results for individual athletes were not always consistent with the “average” finding of
high vs. low angles across tasks for each movement domain.

79
In terms of predicting ACL re-injury, only the knee strategy domain was found
to be statistically significant (Table 7-4, Figure 7-6). Specifically, females who were
assigned to the “high knee extensor bias” cluster (as defined by decreased 2D trunk-
tibia angles across tasks) had a 3.2 higher odds of ACL re-injury compared with those
assigned to the “low knee extensor bias” cluster (as defined by increased 2D trunk-
tibia angles across tasks), after adjusting for potential confounders. For the knee strategy
domain, a decrease in the sagittal plane orientation of the trunk relative to the tibia
(2D trunk-tibia angle) was indicative of decreased reliance on the hip extensors relative
to the knee extensors during the MPA tasks (indicating an increased knee extensor
bias).
143
Our current finding of a high knee extensor biased movement strategy being
predictive of a second ACL injury is consistent with a prior prospective study that
reported females who exhibited greater peak knee extensor moments during a double-
limb drop jump were at increased risk of initial ACL injury.
77

Within the current study, the Likelihood ratio test indicated that a knee strategy
movement tendency improved model fit beyond a baseline model with simply patient
data (Table 7-5). Although our model with the knee strategy domain produced an AUC
of 78% (indicating fair discrimination to distinguish females with vs. without ACL re-
injury), the baseline model had an AUC of 76% (Table 7-5). As such, only a 2%
improvement in AUC occurred with the addition of the movement variable knee
strategy (which was not statistically significant). However, using the improvement in
AUC to assess added predictive ability has been criticized as this approach lacks
statistical power due to its reliance solely on ranks.
21,22,54,112-115
As such, the AUC
improvement test cannot detect important differences in absolute risk, particularly
when the baseline model is already strong.
21,22,54,112-115
Therefore, we also investigated
the IDI index to quantify how inclusion of the knee strategy domain altered the risk
probability for both cases and controls.
21,112-114
On average, inclusion of the knee strategy
domain improved the IDI by 6% (representing a 4% increased risk for cases and 2%
decrease risk for controls), which was statistically significant (Table 7-5). Importantly,

80
17 of 23 cases exhibited a high knee extensor bias movement strategy and experienced
an increased probability of ACL re-injury ranging from 5% to 12%. While a 5% shift
in probability may be deemed clinically unimportant, the fact that the same high-risk
movement strategy could increase risk by 10% or more in several patients (i.e., 11 of
17 cases with a high knee extensor bias) could be relevant.
Apart from the knee strategy domain, no other movement domain was found to
be predictive of ACL re-injury. Particularly interesting is the fact that knee stability
(quantified using the FPPA) was not a significant predictor of ACL re-injury. Although
the FPPA has been proposed to be representative of knee valgus angles and moments,
the validity of this measurement with respect to predicting 3D motion has been
questioned and the ability to predict frontal plane knee moments is somewhat
limited.[Chapter 6] Overall, our findings for the FPPA are in agreement with a recent
systematic review that concluded there is limited evidence for increased frontal plane
knee loading (FPPA, knee valgus angles, or moments) in predicting initial ACL injury.
26

In addition to the knee stability domain, the constructs of trunk stability, pelvis
stability, and shock absorption also were not predictive of ACL re-injury. Although
each of these movement attributes have been shown in prospective studies to be
predictive of initial or secondary ACL injury in various studies,
70,71,77,79,80,119
comparison
of the current findings to existing literature is difficult owing to highly varied approaches
in study design and statistical approaches. For example, some studies have identified
movement risk factors after controlling for non-movement factors that may influence
ACL injury,
77,78,80
while other studies have combined multiple movement risk factors
into a single model (by creating novel metrics or entering them separately).
70,71,119
In
addition, results of predictive studies in this area are difficult to compare owing to
differences in how movement is assessed (2D, 3D, or visual), the sex of the study
sample (females, males, or males/females), and variations in the tasks assessed, just to
name a few.
59,70,71,75,77,79,80,101,110,116,119

81
The results of this study have clinical implications. Most importantly, the
findings highlight the need to expand return to sport assessments post ACLR to include
an assessment of movement attributes, in addition to typically obtained measures of
strength, hop distance performance, etc. The need to redefine return to sport testing
following ACLR is underscored by a recent meta-analysis that reported meeting current
return to sport criteria only decreases risk of injury to the surgical limb while risk to
the contralateral limb actually is increased.
153
Given that ACL re-injury is just as
prevalent in the contralateral limb relative to the ipsilateral limb (and sometimes more
prevalent depending on the follow-up time),
50,160
suggests that assessment of movement
characteristics should be made bilaterally to better inform return to sport decisions.
The fact that the movement construct of knee strategy likely applies to both limbs
suggests that this variable should be considered when assessing an athlete’s preferred
movement strategy. To that end, the 2D trunk-tibia angle is a simple measurement that
can be used to infer whether an athlete displays a high knee extensor bias strategy vs.
low knee extensor bias strategy.
There are several limitations in the current study that need to be considered
when interpreting the results. First, clustering is a data-driven method and will generate
a solution regardless of whether it has any meaningful implications. As such, the results
reported herein for female athletes may not be generalizable to other populations
(including males and those without a history of ACLR). Second, the present study was
retrospective in nature. Therefore, prospective studies are needed to validate the current
findings. Lastly, our clustering quality only was “fair” for each movement domain.
Therefore, it’s possible that elimination of the less important tasks could have improved
cluster quality and therefore the prediction results.

Conclusion
The results of the current study demonstrated that the 2D video-based
measurements obtained from the MPA can be used to distinguish high vs. low-risk

82
movement tendencies in female athletes across several movement domains (knee
strategy, shock absorption, knee stability, trunk stability, and pelvis stability). In
addition, our findings suggest that a “generalized” movement strategy that favors use
of the knee extensors relative to the hip extensors may place females at increased risk
for ACL re-injury. Specifically, a “high knee extensor bias” (as defined by decreased
2D trunk-tibia angles across tasks) was shown to be predictive of ACL re-injury. These
findings should be considered in the development of rehabilitative protocols and return
to sport decisions following ACLR.

83
CHAPTER 8:
SUMMARY AND CONCLUSIONS

Approximately 1 in 4 young athletes who return to a high-risk sport after primary
ACL reconstruction (ACLR) will go on to sustain another ACL injury.
109,156

Interestingly, 60-70% of ACL injuries (initial or secondary) occur as a result of non-
contact mechanisms.
2,51
The fact that movement-related impairments are thought to
underlie non-contact ACL injuries,
59,70,71,77,79,80,110,119
highlights the need to assess
movement behavior to better determine readiness to return to sport. As described in
Chapter 2, the lack of a comprehensive clinical assessment to quantify an athlete’s
readiness to return to sport following ACLR led to the development of the video-based
Movement Performance Assessment (MPA). Although the MPA has been developed
and is being used clinically, it remains unknown if the 2D measures of the MPA
represent/predict 3D measures. In addition, it is not known whether the movement
constructs that comprise the MPA are important for predicting ACL re-injury. This
dissertation sought to answer these questions as a first step in establishing the MPA as
a potential clinical tool to assess an athlete’s readiness to return to sport following
ACLR.
Of the movement domains that comprise the MPA, 3 represent kinematic
variables (knee stability, pelvis stability, and trunk stability) and 2 represent kinetic
variables (shock absorption and knee strategy). The purpose of Aim 1 of this dissertation
(Chapters 3-6) was to determine if the video-based 2D movement variables established
for the MPA are representative of the specific 3D kinematic/kinetic variables related
to ACL injury risk based on prior literature. To achieve this aim, 2D and 3D data were
collected simultaneously while healthy athletes performed the 6 tasks conceptualized
for the MPA (step down, drop jump, lateral shuffle, deceleration, triple hop, and side-
step-cut).

84
In terms of validation of the kinetic construct of knee strategy (Chapter 3), the
2D trunk-tibia angle at peak knee flexion was introduced as a method to approximate
use of the hip extensors relative to the knee extensors (i.e., average hip/knee extensor
moment ratio) during all 6 MPA tasks. It previously has been reported that this 2D
metric could be used to approximate the average hip/knee extensor moment ratio
during squatting,
8
which motivated study of this metric during more comprehensive
and dynamic movements. The results indicated that the 2D trunk-tibia angle could be
used to approximate use of the hip extensors relative to the knee extensors (i.e., average
hip/knee extensor moment ratio) during all 6 MPA tasks, with predictability ranging
from 17% to 77% after adjusting for body mass. More specifically, the more forward
the trunk inclination relative to the tibia, the higher the average hip/knee extensor
moment ratio.
In terms of the validation of the kinetic construct of shock absorption (Chapter
4), the 2D thigh angle at peak knee flexion was introduced as a method to approximate
measures of peak vertical ground reaction force (vGRF) and vGRF impulse during the
MPA tasks that involved impact with the ground. The results indicated that the 2D
thigh angle could be used to approximate impact forces during the relevant MPA tasks,
with predictability ranging from 13% to 47%. More specifically, the lower the 2D thigh
angle (which corresponds to decreased hip and knee flexion), the higher the vGRF
variables examined.
In terms of the validation of the kinematic constructs of pelvis and trunk stability
(Chapters 5), 2D frontal pelvis tilt and 2D trunk lean at peak knee flexion were used
to quantify the corresponding angles in 3D in the relevant MPA tasks (i.e., all except
drop jump). Results indicated that both 2D pelvis tilt and 2D trunk lean can be used
to estimate the corresponding 3D motion during stepping, landing, and change of
direction, with predictability (R
2
) ranging from 29% to 85% (with higher predictability
occurring for the trunk). However, the agreement between 2D and 3D angles was
limited, given wide 95% confidence bounds.

85
In terms of validation of the kinematic construct of knee stability (Chapter 6),
the 2D frontal plane projection angle (FPPA) at peak knee flexion was used to quantify
frontal plane knee loading (i.e., average moment, moment at peak knee flexion, and
peak moment) during all 6 MPA tasks. Results indicated that the FPPA can be used
to approximate frontal plane knee loading knee (i.e., average moment, moment at peak
knee flexion, and/or peak moment), with predictability ranging from 12% to 45%
during all 6 MPA tasks except stepping. More specifically, the larger the FPPA
(indicating medial knee collapse), the higher the frontal plane kinetic variable of
interest (indicating increased knee valgus/decreased knee varus moments).
Following validation of the underlying 2D angular metrics within the MPA
movement domains, the purpose of Chapter 7 was to determine which movement
domains evaluated are relevant for predicting non-contact ACL re-injury (ipsilateral or
contralateral) in female athletes, using a retrospective case-control design. To achieve
this aim, females post ACLR who had previously undergone return to sport testing
using the MPA were surveyed, and cases and controls were subsequently matched based
on age (within 5 years), graft type, sport level, and athletic exposures. Using images
extracted from patient records, the 2D metrics from the MPA were calculated using
the methods outlined in Chapters 3-6. Cluster analysis using the 2D metrics across
tasks was then used to identify subgroups (i.e., clusters) within each of the MPA
movement domains. Clustering results indicated that 2 subgroups for each movement
domain maximized heterogeneity. Interestingly, movement tendencies were consistent
across tasks within each movement domain when using a 2-cluster solution, allowing
the ability to distinguish between female athletes exhibiting high vs. low-risk movement
attributes. As such, low vs. high-risk subgroups were established for each separate MPA
movement domain.
Once high vs. low-risk subgroups were established for each separate MPA
movement domain, the risk of ACL re-injury of female athletes assigned to the high-
risk cluster compared to those assigned to the low-risk cluster was explored. Of the 5

86
MPA movement domains evaluated, only knee strategy was predictive of ACL re-
injury. Compared to the “low knee extensor bias” subgroup (defined by high 2D trunk-
tibia angles across tasks), the odds of ACL re-injury were increased in the “high knee
extensor bias” subgroup (defined by low 2D trunk-tibia angles across tasks) (adjusted
OR = 3.19, 95% CI: 1.02, 9.96, p = 0.05). A receiver operating characteristic curve
showed an area under the curve of 78%, indicating fair prediction accuracy.
Taken together, the results of this dissertation support the use of 2D video
analysis across a wide range of athletic tasks to quantify movement impairments that
have been hypothesized to be associated with elevated risk of ACL re-injury.
Specifically, 2D measures to quantify pelvis stability, trunk stability, knee stability,
shock absorption, and knee strategy can be used as clinical surrogates for more complex
3D measures obtained in a laboratory setting. In terms of predicting ACL re-injury
(ipsilateral or contralateral), only the knee strategy movement domain was found to be
relevant, with female athletes with a high knee extensor bias being at increased risk
compared to those with a low knee extensor bias.

Clinical Implications
The findings of this dissertation have several implications for clinical practice.
Most importantly, the results highlight that assessment of movement quality needs to
be part of the return to sport decision making process, beyond common measures of
strength, hop distance, symptoms/functional limitations, balance, etc. Based on a
recent meta-analysis, passing return to sport criteria post ACLR does not appear to
decrease the risk of subsequent injury.
153
Although risk for ipsilateral re-injury risk was
reduced by 60% when passing return to sport tests, there was a 235% increased risk of
injury to the contralateral limb.
153
Interestingly, only 1 of the 5 studies used for these
meta-analysis calculations evaluated movement as part of their return to sport criteria.
This dissertation adds to the growing body of literature that suggests that biomechanical

87
variables (kinematics and/or kinetics) likely contribute to increased ACL re-injury
risk.
70,71,110,119

This dissertation has demonstrated that 2D video can provide clinicians with
reasonable estimates of 3D variables thought to be related to ACL injury and other
clinical conditions not examined in this dissertation. Findings also highlight that 2D
video can be used to determine the athletes with at risk movement tendencies, which
may be useful for injury prevention screening. The most important aspect of the
validation findings was related to the constructs of knee strategy and shock absorption,
as novel 2D metrics not previously described in the literature were introduced. For the
knee strategy domain, it was determined that the 2D trunk-tibia angle can be used to
quantify use of the hip extensors relative to the knee extensive across a wide range of
tasks (stepping, landing, and change of direction). Given that females have a tendency
to exhibit high knee extensor moments relative to hip extensor moments during landing
compared with males,
141
and that overuse of the knee extensors has been suggested as
a risk factor for ACL injury in females based on prospective studies,
77
the results for
the knee strategy construct are clinically important. For shock absorption, it was
determined that the 2D thigh angle can be used to quantify impact forces, which is
clinically relevant given that high impact forces have been implicated in various lower-
extremity injuries (e.g., patellar tendinopathy, ankle instability, and ACL injury).
30,44,78

Among the 5 MPA movement domains, knee strategy emerged as being most
important for predicting ACL re-injury (ipsilateral or contralateral) in female athletes.
Given that overuse of the knee extensors is a risk factor for ACL injury,
141
clinicians
should consider retraining of movement strategies to reduce reliance on the quadriceps
during athletic tasks. This could be achieved by emphasizing forward motion of the
trunk relative to the tibia (which decreases knee extensor bias, or the hip/knee extensor
moment ratio). The fact that females with increased hip/knee extensor strength have
been reported to exhibit increased hip/knee extensor moments during landing,
141

suggests that increasing forward position of the trunk relative to the tibia during

88
dynamic activities may be facilitated by improving strength of the hip extensors relative
to the knee extensors. In addition, motor skill training, as opposed to movement
repetition (e.g., isolated strength training), may be a useful adjunct to train patients
into more protective movement patterns.
120
Benefits of ACL prevention programs may
be the result of neuroplastic changes in the brain due to skill acquisition and motor
learning.
120

Although this dissertation only looked at risk factors for ACL re-injury, the
results may have implications for prevention of the initial injury. Based on a 12-year
prospective intervention study, implementation of a hip-focused training program (with
jump-landing maneuvers, hip strengthening, and balance exercises) significantly
reduced ACL injury incidence in female athletes (i.e., 62% rate reduction of initial
ACL injury).
102
Therefore, it is reasonable to assume that the retraining of movement
patterns to emphasize use of the hip extensors may be protective against initial ACL
injury.

Limitations and Direction for Future Research
This dissertation represented the first step in establishing the clinical utility of
the MPA for ACL re-injury assessment. Future studies are needed to produce a “clinic
friendly” tool that can be used for ACL re-injury movement risk assessment. It is
important that the findings of this dissertation are translated into clinical practice. One
of the most important directions for future research is to externally validate the injury
prediction model findings using a prospective study with a large sample size. The small
number of patients and the retrospective nature of the design was a limitation of the
injury prediction modeling outcomes.
The results indicated that only 17 of 23 female cases in the injury prediction
model demonstrated a “high knee extensor bias.” Therefore, the knee strategy MPA
domain alone was not capable of perfect classification for cases, which was reflected in
the area under the curve (AUC) of 78% (fair). It is possible that various “at-risk”

89
movement profiles may exist that combine different levels of the MPA domains, which
could lead to a more powerful injury prediction model. In other words, the clustering
results produced 10 separate patient groups (2 subgroups for each of the 5 MPA
movement domains), so how the various high vs. low-risk subgroups overlap remains
unknown. As such, further analyses are needed that examine the movement domains
in combination.
Although the injury prediction model findings identified the knee strategy
movement domain as the most relevant for ACL re-injury prediction in female athletes,
it remains unknown which tasks are most important in distinguishing between a high
vs. low knee extensor bias. Therefore, further study to rank the importance of variables
used during clustering is required. This would help clinicians determine which tasks are
most important to emphasize during rehabilitation post ACLR.
Once task importance is identified for the tasks within the knee strategy domain,
clinical cutoffs to distinguish between a high vs. low knee extensor bias across tasks
should be established. Although the average 2D trunk-tibia angles were reported for
both subgroups (Table 7-3), clinically relevant cut-off values were not established within
this dissertation. Clinical cutoffs could be established for each task independently.
Alternatively, clinical cutoffs could be established using only the most important tasks
in combination, which would produce a more parsimonious model. It is possible that
only a subset of tasks is needed to determine an athlete’s “generalized” movement
strategy (i.e., high vs low knee extensor bias).
The injury prediction model was specific to female athletes only. Further study
is required to establish a predictive model suitable for males. Although males and
females have similar ACL re-injury risk,
107
it is well established that males and females
display differing movement profiles based on cross-sectional studies.
15,127,137,141

Because of a limited sample size, the injury prediction models were developed
by combining data from both the ipsilateral and contralateral limbs. However,
asymmetrical lower-extremity loading (such as decreased knee extensor moments on

90
the surgical side) has been reported post ACLR during various activities.
43,103,108,129,136

Although loading asymmetries may normalize with time, asymmetries can persist 2
years or more after ACLR.
63,108,133
In addition, a recent meta-analysis reported that
although passing return to sport test batteries for the surgical limb reduced ipsilateral
risk re-injury risk by 60%, there was a 235% increased risk of injury to the contralateral
limb.
153
Therefore, further study is warranted to explore whether the 2D trunk-tibia
angle is indeed consistent between limbs.
Prior to building an injury prediction model, cluster analysis was used to form
subgroups across tasks for each movement domain. The underlying 2D metrics within
each movement domain were consistent in direction when using a 2-cluster solution,
indicating that a definable clinical movement profile exists (high vs. low) for each MPA
movement domain in patients post ACLR. Whether a similar definable movement
profile exists in healthy patients remains unknown and should be the focus of future
research, which would help inform development of the MPA for risk assessment of
initial ACL injury (or other lower-extremity injuries).
Although the current dissertation established intra-rater reliability and inter-rater
reliability for all the MPA metrics, only two raters were assessed. Therefore, more
formal reliability studies are needed, particularly with raters with no prior experience
using the MPA. In addition, the stability of patients’ test scores when the MPA is
administrated at different time points remains unknown. As such, test re-test reliability
needs to be established before the MPA can be recommended in clinical practice.

Conclusion
The results of this dissertation support the use of 2D video analysis across a wide
range of athletic tasks to quantify movement impairments that have been hypothesized
to be associated with elevated risk of ACL re-injury. Specifically, 2D measures to
quantify pelvis stability, trunk stability, knee stability, shock absorption, and knee
strategy can be used as clinical surrogates for more complex 3D measures obtained in

91
a laboratory setting. In terms of predicting ACL re-injury (ipsilateral or contralateral),
only the knee strategy domain was found to be relevant, with female athletes with a
high knee extensor bias being at increased risk compared to those with a low knee
extensor bias. Future studies are now needed to verify these injury prediction results in
prospective studies, establish task importance for quantifying knee strategy, and
determine cutoffs that can be used in the clinical setting to distinguish between a high
vs. low knee extensor bias.

92
BIBLIOGRAPHY:

1. Ageberg E, Bennell KL, Hunt MA, et al. Validity and inter-rater reliability of medio-
lateral knee motion observed during a single-limb mini squat. BMC Musculoskelet
Disord. 2010;11:265.
2. Agel J, Rockwood T, Klossner D. Collegiate ACL Injury Rates Across 15 Sports:
National Collegiate Athletic Association Injury Surveillance System Data Update (2004-
2005 Through 2012-2013). Clin J Sport Med. 2016;26(6):518-523.
3. Alahmari A, Herrington L, Jones R. Concurrent validity of two-dimensional video
analysis of lower-extremity frontal plane of movement during multidirectional single-leg
landing. Phys Ther Sport. 2020;42:40-45.
4. Bahr R, Krosshaug T. Understanding injury mechanisms: a key component of preventing
injuries in sport. Br J Sports Med. 2005;39(6):324-329.
5. Ball JR, Harris CB, Lee J, Vives MJ. Lumbar Spine Injuries in Sports: Review of the
Literature and Current Treatment Recommendations. Sports Med Open. 2019;5(1):26.
6. Barber-Westin SD, Noyes FR. Factors used to determine return to unrestricted sports
activities after anterior cruciate ligament reconstruction. Arthroscopy. 2011;27(12):1697-
1705.
7. Barrack AJ, Straub RK, Cannon J, Powers CM. the relative orientation of the trunk and
tibia can be used to estimate the demands on the hip and knee extensors during the
barbell back squat. International Journal of Sports Science & Coaching.
2021:1747954121997957.
8. Barrack AJ, Straub RK, Cannon J, Powers CM. The relative orientation of the trunk and
tibia can be used to estimate the demands on the hip and knee extensors during the
barbell back squat. Int J Sports Sci Coaching. 2021;16(4):1004-1010
9. Bates NA, Schilaty ND, Nagelli CV, Krych AJ, Hewett TE. Novel mechanical impact
simulator designed to generate clinically relevant anterior cruciate ligament ruptures. Clin
Biomech (Bristol, Avon). 2017;44:36-44.

93
10. Belyea BC, Lewis E, Gabor Z, Jackson J, King DL. Validity and Intrarater Reliability of
2-Dimensional Motion Analysis Using a Handheld Tablet Compared to Traditional 3-
Dimensional Motion Analysis. J Sport Rehabil. 2015;24(4).
11. Biscarini A, Botti FM, Pettorossi VE. Joint torques and joint reaction forces during
squatting with a forward or backward inclined Smith machine. J Appl Biomech.
2013;29(1):85-97.
12. Blackburn JT, Padua DA. Sagittal-plane trunk position, landing forces, and quadriceps
electromyographic activity. J Athl Train. 2009;44(2):174-179.
13. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods
of clinical measurement. Lancet. 1986;1(8476):307-310.
14. Boden BP, Dean GS, Feagin JA, Jr., Garrett WE, Jr. Mechanisms of anterior cruciate
ligament injury. Orthopedics. 2000;23(6):573-578.
15. Carson DW, Ford KR. Sex differences in knee abduction during landing: a systematic
review. Sports Health. 2011;3(4):373-382.
16. Chappell JD, Yu B, Kirkendall DT, Garrett WE. A comparison of knee kinetics between
male and female recreational athletes in stop-jump tasks. Am J Sports Med.
2002;30(2):261-267.
17. Chijimatsu M, Ishida T, Yamanaka M, et al. Landing instructions focused on pelvic and
trunk lateral tilt decrease the knee abduction moment during a single-leg drop vertical
jump. Phys Ther Sport. 2020;46:226-233.
18. Cohen J. A power primer. Psychol Bull. 1992;112(1):155.
19. Cohen J. Statistical power analysis for the behavioral sciences. Academic press; 2013.
20. Cook G, Burton L, Hoogenboom BJ, Voight M. Functional movement screening: the use
of fundamental movements as an assessment of function - part 1. Int J Sports Phys Ther.
2014;9(3):396-409.
21. Cook NR. Quantifying the added value of new biomarkers: how and how not. Diagn
Progn Res. 2018;2:14.

94
22. Cook NR. Use and misuse of the receiver operating characteristic curve in risk
prediction. Circulation. 2007;115(7):928-935.
23. Corry II, Webb J. Injuries of the sporting knee. Br J Sports Med. 2000;34(5):395.
24. Cortes N, Onate J, Van Lunen B. Pivot task increases knee frontal plane loading
compared with sidestep and drop-jump. J Sports Sci. 2011;29(1):83-92.
25. Cowan DN, Bedno SA, Urban N, Lee DS, Niebuhr DW. Step test performance and risk
of stress fractures among female army trainees. Am J Prev Med. 2012;42(6):620-624.
26. Cronstrom A, Creaby MW, Ageberg E. Do knee abduction kinematics and kinetics
predict future anterior cruciate ligament injury risk? A systematic review and meta-
analysis of prospective studies. BMC Musculoskelet Disord. 2020;21(1):563.
27. Crossley KM, Zhang WJ, Schache AG, Bryant A, Cowan SM. Performance on the
single-leg squat task indicates hip abductor muscle function. Am J Sports Med.
2011;39(4):866-873.
28. Dai B, Butler RJ, Garrett WE, Queen RM. Using ground reaction force to predict knee
kinetic asymmetry following anterior cruciate ligament reconstruction. Scand J Med Sci
Sports. 2014;24(6):974-981.
29. Dai B, Herman D, Liu H, Garrett WE, Yu B. Prevention of ACL injury, part I: injury
characteristics, risk factors, and loading mechanism. Res Sports Med. 2012;20(3-4):180-
197.
30. Dayakidis MK, Boudolos K. Ground reaction force data in functional ankle instability
during two cutting movements. Clin Biomech (Bristol, Avon). 2006;21(4):405-411.
31. Decker MJ, Torry MR, Wyland DJ, Sterett WI, Richard Steadman J. Gender differences
in lower extremity kinematics, kinetics and energy absorption during landing. Clin
Biomech (Bristol, Avon). 2003;18(7):662-669.
32. DeHaven KE, Lintner DM. Athletic injuries: comparison by age, sport, and gender. Am J
Sports Med. 1986;14(3):218-224.

95
33. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more
correlated receiver operating characteristic curves: a nonparametric approach. Biometrics.
1988;44(3):837-845.
34. Devita P, Skelly WA. Effect of landing stiffness on joint kinetics and energetics in the
lower extremity. Med Sci Sports Exerc. 1992;24(1):108-115.
35. Dingenen B, Malfait B, Nijs S, et al. Postural Stability During Single-Leg Stance: A
Preliminary Evaluation of Noncontact Lower Extremity Injury Risk. J Orthop Sports
Phys Ther. 2016;46(8):650-657.
36. Dingenen B, Malfait B, Nijs S, et al. Can two-dimensional video analysis during single-
leg drop vertical jumps help identify non-contact knee injury risk? A one-year
prospective study. Clin Biomech 2015;30(8):781-787.
37. Dingenen B, Staes FF, Santermans L, et al. Are two-dimensional measured frontal plane
angles related to three-dimensional measured kinematic profiles during running? Phys
Ther Sport. 2018;29:84-92.
38. Dos'Santos T, McBurnie A, Donelon T, et al. A qualitative screening tool to identify
athletes with 'high-risk' movement mechanics during cutting: The cutting movement
assessment score (CMAS). Phys Ther Sport. 2019;38:152-161.
39. DuPrey KM, Liu K, Cronholm PF, et al. Baseline Time to Stabilization Identifies
Anterior Cruciate Ligament Rupture Risk in Collegiate Athletes. Am J Sports Med.
2016;44(6):1487-1491.
40. Edwards S, Steele JR, Cook JL, et al. Characterizing patellar tendon loading during the
landing phases of a stop-jump task. Scand J Med Sci Sports. 2012;22(1):2-11.
41. Eime RM, Young JA, Harvey JT, Charity MJ, Payne WR. A systematic review of the
psychological and social benefits of participation in sport for adults: informing
development of a conceptual model of health through sport. Int J Behav Nutr Phys Act.
2013;10:135.
42. Eime RM, Young JA, Harvey JT, Charity MJ, Payne WR. A systematic review of the
psychological and social benefits of participation in sport for children and adolescents:
informing development of a conceptual model of health through sport. Int J Behav Nutr
Phys Act. 2013;10:98.

96
43. Ernst GP, Saliba E, Diduch DR, Hurwitz SR, Ball DW. Lower extremity compensations
following anterior cruciate ligament reconstruction. Phys Ther. 2000;80(3):251-260.
44. Fietzer AL, Chang YJ, Kulig K. Dancers with patellar tendinopathy exhibit higher
vertical and braking ground reaction forces during landing. J Sports Sci.
2012;30(11):1157-1163.
45. Formann AK. Die latent-class-analyse: Einführung in Theorie und Anwendung. Beltz;
1984.
46. Fry AC, Smith JC, Schilling BK. Effect of knee position on hip and knee torques during
the barbell squat. J Strength Cond Res. 2003;17(4):629-633.
47. Gail MH, Haneuse S. Power and sample size for multivariate logistic modeling of
unmatched case-control studies. Stat Methods Med Res. 2019;28(3):822-834.
48. Gail MH, Wheeler MW. Package ‘samplesizelogisticcasecontrol’. 2017.
49. Gornitzky AL, Lott A, Yellin JL, et al. Sport-Specific Yearly Risk and Incidence of
Anterior Cruciate Ligament Tears in High School Athletes: A Systematic Review and
Meta-analysis. Am J Sports Med. 2016;44(10):2716-2723.
50. Grassi A, Macchiarola L, Lucidi GA, et al. More Than a 2-Fold Risk of Contralateral
Anterior Cruciate Ligament Injuries Compared With Ipsilateral Graft Failure 10 Years
After Primary Reconstruction. Am J Sports Med. 2020;48(2):310-317.
51. Group M, Ding DY, Zhang AL, et al. Subsequent Surgery After Revision Anterior
Cruciate Ligament Reconstruction: Rates and Risk Factors From a Multicenter Cohort.
Am J Sports Med. 2017;45(9):2068-2076.
52. Gwynne CR, Curran SA. Two-dimensional frontal plane projection angle can identify
subgroups of patellofemoral pain patients who demonstrate dynamic knee valgus. Clin
Biomech (Bristol, Avon). 2018;58:44-48.
53. Haitz K, Shultz R, Hodgins M, Matheson GO. Test-retest and interrater reliability of the
functional lower extremity evaluation. J Orthop Sports Phys Ther. 2014;44(12):947-954.

97
54. Harrell FE. Regression modeling strategies: with applications to linear models, logistic
and ordinal regression, and survival analysis. Vol 3: Springer; 2015.
55. Havens KL, Sigward SM. Joint and segmental mechanics differ between cutting
maneuvers in skilled athletes. Gait Posture. 2015;41(1):33-38.
56. Havlicek LL, Peterson NL. Robustness of the Pearson correlation against violations of
assumptions. Percept Mot Skills. 1976;43(3_suppl):1319-1334.
57. Herrington L. Knee valgus angle during single leg squat and landing in patellofemoral
pain patients and controls. Knee. 2014;21(2):514-517.
58. Herrington L, Alenezi F, Alzhrani M, Alrayani H, Jones R. The reliability and criterion
validity of 2D video assessment of single leg squat and hop landing. J Electromyogr
Kinesiol. 2017;34:80-85.
59. Hewett TE, Myer GD, Ford KR, et al. Biomechanical measures of neuromuscular control
and valgus loading of the knee predict anterior cruciate ligament injury risk in female
athletes: a prospective study. Am J Sports Med. 2005;33(4):492-501.
60. Hewett TE, Torg JS, Boden BP. Video analysis of trunk and knee motion during non-
contact anterior cruciate ligament injury in female athletes: lateral trunk and knee
abduction motion are combined components of the injury mechanism. Br J Sports Med.
2009;43(6):417-422.
61. Hopkins WG. Measures of reliability in sports medicine and science. Sports Med.
2000;30(1):1-15.
62. Hubert L, Köhn H-F, Steinley D. Cluster analysis: a toolbox for MATLAB. The SAGE
handbook of quantitative methods in psychology. 2009:444-512.
63. Hughes G, Musco P, Caine S, Howe L. Lower Limb Asymmetry After Anterior Cruciate
Ligament Reconstruction in Adolescent Athletes: A Systematic Review and Meta-
Analysis. J Athl Train. 2020;55(8):811-825.
64. Janse van Rensburg L, Dare M, Louw Q, et al. Pelvic and hip kinematics during single-
leg drop-landing are altered in sports participants with long-standing groin pain: A cross-
sectional study. Phys Ther Sport. 2017;26:20-26.

98
65. Joseph AM, Collins CL, Henke NM, et al. A multisport epidemiologic comparison of
anterior cruciate ligament injuries in high school athletics. J Athl Train. 2013;48(6):810-
817.
66. Kaufman L, Rousseeuw PJ. Finding groups in data: an introduction to cluster analysis.
John Wiley & Sons; 2009.
67. Kay MC, Register-Mihalik JK, Gray AD, et al. The Epidemiology of Severe Injuries
Sustained by National Collegiate Athletic Association Student-Athletes, 2009-2010
Through 2014-2015. J Athl Train. 2017;52(2):117-128.
68. Kernozek TW, Gheidi N, Zellmer M, et al. Effects of Anterior Knee Displacement
During Squatting on Patellofemoral Joint Stress. J Sport Rehabil. 2018;27(3):237-243.
69. Khayambashi K, Ghoddosi N, Straub RK, Powers CM. Hip Muscle Strength Predicts
Noncontact Anterior Cruciate Ligament Injury in Male and Female Athletes: A
Prospective Study. Am J Sports Med. 2016;44(2):355-361.
70. King E, Richter C, Daniels KAJ, et al. Biomechanical but Not Strength or Performance
Measures Differentiate Male Athletes Who Experience ACL Reinjury on Return to Level
1 Sports. Am J Sports Med. 2021;49(4):918-927.
71. King E, Richter C, Daniels KAJ, et al. Can Biomechanical Testing After Anterior
Cruciate Ligament Reconstruction Identify Athletes at Risk for Subsequent ACL Injury
to the Contralateral Uninjured Limb? Am J Sports Med. 2021;49(3):609-619.
72. Kingston B, Murray A, Norte GE, Glaviano NR. Validity and reliability of 2-dimensional
trunk, hip, and knee frontal plane kinematics during single-leg squat, drop jump, and
single-leg hop in females with patellofemoral pain. Phys Ther Sport. 2020;45:181-187.
73. Koga H, Nakamae A, Shima Y, et al. Mechanisms for noncontact anterior cruciate
ligament injuries: knee joint kinematics in 10 injury situations from female team handball
and basketball. Am J Sports Med. 2010;38(11):2218-2225.
74. Krosshaug T, Nakamae A, Boden BP, et al. Mechanisms of anterior cruciate ligament
injury in basketball: video analysis of 39 cases. Am J Sports Med. 2007;35(3):359-367.

99
75. Krosshaug T, Steffen K, Kristianslund E, et al. The Vertical Drop Jump Is a Poor
Screening Test for ACL Injuries in Female Elite Soccer and Handball Players: A
Prospective Cohort Study of 710 Athletes. Am J Sports Med. 2016;44(4):874-883.
76. Laughlin WA, Weinhandl JT, Kernozek TW, et al. The effects of single-leg landing
technique on ACL loading. J Biomech. 2011;44(10):1845-1851.
77. Leppanen M, Pasanen K, Krosshaug T, et al. Sagittal Plane Hip, Knee, and Ankle
Biomechanics and the Risk of Anterior Cruciate Ligament Injury: A Prospective Study.
Orthop J Sports Med. 2017;5(12):2325967117745487.
78. Leppanen M, Pasanen K, Kujala UM, et al. Stiff Landings Are Associated With
Increased ACL Injury Risk in Young Female Basketball and Floorball Players. Am J
Sports Med. 2017;45(2):386-393.
79. Leppanen M, Pasanen K, Kujala UM, et al. Stiff Landings Are Associated With
Increased ACL Injury Risk in Young Female Basketball and Floorball Players: Response.
Am J Sports Med. 2017;45(3):NP5-NP6.
80. Leppanen M, Rossi MT, Parkkari J, et al. Altered hip control during a standing knee-lift
test is associated with increased risk of knee injuries. Scand J Med Sci Sports.
2020;30(5):922-931.
81. Lewis CL, Foch E, Luko MM, Loverro KL, Khuu A. Differences in Lower Extremity and
Trunk Kinematics between Single Leg Squat and Step Down Tasks. PLoS One.
2015;10(5):e0126258.
82. Lletı R, Ortiz MC, Sarabia LA, Sánchez MS. Selecting variables for k-means cluster
analysis by using a genetic algorithm that optimises the silhouettes. Anal Chim Acta.
2004;515(1):87-100.
83. Lopes TJA, Ferrari D, Ioannidis J, et al. Reliability and Validity of Frontal Plane
Kinematics of the Trunk and Lower Extremity Measured With 2-Dimensional Cameras
During Athletic Tasks: A Systematic Review With Meta-analysis. J Orthop Sports Phys
Ther. 2018;48(10):812-822.
84. Lorenzetti S, Gulay T, Stoop M, et al. Comparison of the angles and corresponding
moments in the knee and hip during restricted and unrestricted squats. J Strength Cond
Res. 2012;26(10):2829-2836.

100
85. Maykut JN, Taylor-Haas JA, Paterno MV, DiCesare CA, Ford KR. Concurrent validity
and reliability of 2d kinematic analysis of frontal plane motion during running. Int J
Sports Phys Ther. 2015;10(2):136-146.
86. McCunn R, Aus der Funten K, Fullagar HH, McKeown I, Meyer T. Reliability and
Association with Injury of Movement Screens: A Critical Review. Sports Med.
2016;46(6):763-781.
87. McCunn R, Aus der Funten K, Govus A, et al. The Intra- and Inter-Rater Reliability of
the Soccer Injury Movement Screen (Sims). Int J Sports Phys Ther. 2017;12(1):53-66.
88. McCunn R, Aus der Funten K, Whalan M, Sampson JA, Meyer T. Soccer Injury
Movement Screen (SIMS) Composite Score Is Not Associated With Injury Among
Semiprofessional Soccer Players. J Orthop Sports Phys Ther. 2018;48(8):630-636.
89. McKeown I, Taylor-McKeown K, Woods C, Ball N. Athletic ability assessment: a
movement assessment protocol for athletes. Int J Sports Phys Ther. 2014;9(7):862-873.
90. McLean SG, Walker K, Ford KR, et al. Evaluation of a two dimensional analysis method
as a screening and evaluation tool for anterior cruciate ligament injury. Br J Sports Med.
2005;39(6):355-362.
91. Milner CE, Fairbrother JT, Srivatsan A, Zhang S. Simple verbal instruction improves
knee biomechanics during landing in female athletes. Knee. 2012;19(4):399-403.
92. Mineta S, Inami T, Hoshiba T, et al. Greater knee varus angle and pelvic internal rotation
after landing are predictive factors of a non-contact lateral ankle sprain. Phys Ther Sport.
2021;50:59-64.
93. Mizner RL, Chmielewski TL, Toepke JJ, Tofte KB. Comparison of 2-dimensional
measurement techniques for predicting knee angle and moment during a drop vertical
jump. Clin J Sport Med. 2012;22(3):221-227.
94. Moran RW, Schneiders AG, Mason J, Sullivan SJ. Do Functional Movement Screen
(FMS) composite scores predict subsequent injury? A systematic review with meta-
analysis. Br J Sports Med. 2017;51(23):1661-1669.
95. Mukaka MM. Statistics corner: A guide to appropriate use of correlation coefficient in
medical research. Malawi Med J. 2012;24(3):69-71.

101
96. Myer GD, Ford KR, Barber Foss KD, et al. The incidence and potential pathomechanics
of patellofemoral pain in female athletes. Clin Biomech (Bristol, Avon). 2010;25(7):700-
707.
97. Myer GD, Ford KR, Hewett TE. Tuck Jump Assessment for Reducing Anterior Cruciate
Ligament Injury Risk. Athl Ther Today. 2008;13(5):39-44.
98. Nakagawa TH, Moriya ET, Maciel CD, Serrao AF. Frontal plane biomechanics in males
and females with and without patellofemoral pain. Med Sci Sports Exerc.
2012;44(9):1747-1755.
99. Ness BM, Taylor AL, Haberl MD, Reuteman PF, Borgert AJ. Clinical observation and
analysis of movement quality during performance on the star excursion balance test. Int J
Sports Phys Ther. 2015;10(2):168-177.
100. Nilstad A, Andersen TE, Kristianslund E, et al. Physiotherapists can identify female
football players with high knee valgus angles during vertical drop jumps using real-time
observational screening. J Orthop Sports Phys Ther. 2014;44(5):358-365.
101. Nilstad A, Petushek E, Mok KM, Bahr R, Krosshaug T. Kiss goodbye to the 'kissing
knees': no association between frontal plane inward knee motion and risk of future non-
contact ACL injury in elite female athletes. Sports Biomech. 2021:1-15.
102. Omi Y, Sugimoto D, Kuriyama S, et al. Effect of Hip-Focused Injury Prevention
Training for Anterior Cruciate Ligament Injury Reduction in Female Basketball Players:
A 12-Year Prospective Intervention Study. Am J Sports Med. 2018;46(4):852-861.
103. Orishimo KF, Kremenic IJ, Mullaney MJ, McHugh MP, Nicholas SJ. Adaptations in
single-leg hop biomechanics following anterior cruciate ligament reconstruction. Knee
Surg Sports Traumatol Arthrosc. 2010;18(11):1587-1593.
104. Padua DA, DiStefano LJ, Beutler AI, et al. The Landing Error Scoring System as a
Screening Tool for an Anterior Cruciate Ligament Injury-Prevention Program in Elite-
Youth Soccer Athletes. J Athl Train. 2015;50(6):589-595.
105. Padua DA, Marshall SW, Boling MC, et al. The Landing Error Scoring System (LESS) Is
a valid and reliable clinical assessment tool of jump-landing biomechanics: The JUMP-
ACL study. In: Am J Sports Med. Vol 37. United States2009:1996-2002.

102
106. Parsonage JR, Williams RS, Rainer P, McKeown I, Williams MD. Assessment of
conditioning-specific movement tasks and physical fitness measures in talent identified
under 16-year-old rugby union players. J Strength Cond Res. 2014;28(6):1497-1506.
107. Patel AD, Bullock GS, Wrigley J, et al. Does sex affect second ACL injury risk? A
systematic review with meta-analysis. Br J Sports Med. 2021;55(15):873-882.
108. Paterno MV, Ford KR, Myer GD, Heyl R, Hewett TE. Limb asymmetries in landing and
jumping 2 years following anterior cruciate ligament reconstruction. Clin J Sport Med.
2007;17(4):258-262.
109. Paterno MV, Rauh MJ, Schmitt LC, Ford KR, Hewett TE. Incidence of contralateral and
ipsilateral anterior cruciate ligament (ACL) injury after primary ACL reconstruction and
return to sport. Clin J Sport Med. 2012;22(2):116-121.
110. Paterno MV, Schmitt LC, Ford KR, et al. Biomechanical measures during landing and
postural stability predict second anterior cruciate ligament injury after anterior cruciate
ligament reconstruction and return to sport. Am J Sports Med. 2010;38(10):1968-1978.
111. Pearce N. Analysis of matched case-control studies. BMJ. 2016;352:i969.
112. Pencina MJ, D'Agostino RB, Pencina KM, Janssens AC, Greenland P. Interpreting
incremental value of markers added to risk prediction models. Am J Epidemiol.
2012;176(6):473-481.
113. Pencina MJ, D'Agostino RB, Sr., D'Agostino RB, Jr., Vasan RS. Evaluating the added
predictive ability of a new marker: from area under the ROC curve to reclassification and
beyond. Stat Med. 2008;27(2):157-172; discussion 207-112.
114. Pencina MJ, Parikh CR, Kimmel PL, et al. Statistical methods for building better
biomarkers of chronic kidney disease. Stat Med. 2019;38(11):1903-1917.
115. Pepe MS, Kerr KF, Longton G, Wang Z. Testing for improvement in prediction model
performance. Stat Med. 2013;32(9):1467-1482.
116. Petushek E, Nilstad A, Bahr R, Krosshaug T. Drop Jump? Single-Leg Squat? Not if You
Aim to Predict Anterior Cruciate Ligament Injury From Real-Time Clinical Assessment:
A Prospective Cohort Study Involving 880 Elite Female Athletes. J Orthop Sports Phys
Ther. 2021;51(7):372-378.

103
117. Plisky PJ, Rauh MJ, Kaminski TW, Underwood FB. Star Excursion Balance Test as a
predictor of lower extremity injury in high school basketball players. J Orthop Sports
Phys Ther. 2006;36(12):911-919.
118. Pollard CD, Sigward SM, Powers CM. Limited hip and knee flexion during landing is
associated with increased frontal plane knee motion and moments. Clin Biomech (Bristol,
Avon). 2010;25(2):142-146.
119. Poston GR, Schmitt LC, Ithurburn MP, et al. Reduced 2-D Frontal Plane Motion During
Single-Limb Landing Is Associated With Risk of Future Anterior Cruciate Ligament
Graft Rupture After Anterior Cruciate Ligament Reconstruction and Return to Sport: A
Pilot Study. J Orthop Sports Phys Ther. 2021;51(2):82-87.
120. Powers CM, Fisher B. Mechanisms underlying ACL injury-prevention training: the
brain-behavior relationship. J Athl Train. 2010;45(5):513-515.
121. Rabin A, Portnoy S, Kozol Z. The Association Between Visual Assessment of Quality of
Movement and Three-Dimensional Analysis of Pelvis, Hip, and Knee Kinematics During
a Lateral Step Down Test. J Strength Cond Res. 2016;30(11):3204-3211.
122. Raisanen AM, Pasanen K, Krosshaug T, et al. Association between frontal plane knee
control and lower extremity injuries: a prospective study on young team sport athletes.
BMJ Open Sport Exerc Med. 2018;4(1):e000311.
123. Reid DA, Vanweerd RJ, Larmer PJ, Kingstone R. The inter and intra rater reliability of
the Netball Movement Screening Tool. J Sci Med Sport. 2015;18(3):353-357.
124. Renstrom P, Ljungqvist A, Arendt E, et al. Non-contact ACL injuries in female athletes:
an International Olympic Committee current concepts statement. Br J Sports Med.
2008;42(6):394-412.
125. Rossi MK, Pasanen K, Heinonen A, et al. Performance in dynamic movement tasks and
occurrence of low back pain in youth floorball and basketball players. BMC
Musculoskelet Disord. 2020;21(1):350.
126. Rowley KM, Richards JG. Increasing plantarflexion angle during landing reduces vertical
ground reaction forces, loading rates and the hip's contribution to support moment within
participants. J Sports Sci. 2015;33(18):1922-1931.

104
127. Russell KA, Palmieri RM, Zinder SM, Ingersoll CD. Sex differences in valgus knee
angle during a single-leg drop jump. J Athl Train. 2006;41(2):166-171.
128. Ryman Augustsson S, Ageberg E. Weaker lower extremity muscle strength predicts
traumatic knee injury in youth female but not male athletes. BMJ Open Sport Exerc Med.
2017;3(1):e000222.
129. Salem GJ, Salinas R, Harding FV. Bilateral kinematic and kinetic analysis of the squat
exercise after anterior cruciate ligament reconstruction. Arch Phys Med Rehabil.
2003;84(8):1211-1216.
130. Scholtes SA, Salsich GB. A Dynamic Valgus Index That Combines Hip and Knee
Angles: Assessment of Utility in Females with Patellofemoral Pain. Int J Sports Phys
Ther. 2017;12(3):333-340.
131. Schurr SA, Marshall AN, Resch JE, Saliba SA. Two-Dimensional Video Analysis Is
Comparable to 3d Motion Capture in Lower Extremity Movement Assessment. Int J
Sports Phys Ther. 2017;12(2):163-172.
132. Sell TC, Ferris CM, Abt JP, et al. Predictors of proximal tibia anterior shear force during
a vertical stop-jump. J Orthop Res. 2007;25(12):1589-1597.
133. Sharafoddin-Shirazi F, Letafatkar A, Hogg J, Saatchian V. Biomechanical asymmetries
persist after ACL reconstruction: results of a 2-year study. J Exp Orthop. 2020;7(1):86.
134. Shimokochi Y, Shultz SJ. Mechanisms of noncontact anterior cruciate ligament injury. J
Athl Train. 2008;43(4):396-408.
135. Sigward SM, Cesar GM, Havens KL. Predictors of Frontal Plane Knee Moments During
Side-Step Cutting to 45 and 110 Degrees in Men and Women: Implications for Anterior
Cruciate Ligament Injury. Clin J Sport Med. 2015;25(6):529-534.
136. Sigward SM, Lin P, Pratt K. Knee loading asymmetries during gait and running in early
rehabilitation following anterior cruciate ligament reconstruction: A longitudinal study.
Clin Biomech (Bristol, Avon). 2016;32:249-254.
137. Sigward SM, Powers CM. The influence of gender on knee kinematics, kinetics and
muscle activation patterns during side-step cutting. Clin Biomech (Bristol, Avon).
2006;21(1):41-48.

105
138. Simon M, Parizek C, Earl-Boehm JE, Bazett-Jones DM. Quantitative and qualitative
assessment of frontal plane knee motion in males and females: A reliability and validity
study. Knee. 2018;25(6):1057-1064.
139. Slater A, Campbell A, Smith A, Straker L. Greater lower limb flexion in gymnastic
landings is associated with reduced landing force: a repeated measures study. Sports
Biomech. 2015;14(1):45-56.
140. Smith HC, Johnson RJ, Shultz SJ, et al. A prospective evaluation of the Landing Error
Scoring System (LESS) as a screening tool for anterior cruciate ligament injury risk. In:
Am J Sports Med. Vol 40.2012:521-526.
141. Stearns KM, Keim RG, Powers CM. Influence of relative hip and knee extensor muscle
strength on landing biomechanics. Med Sci Sports Exerc. 2013;45(5):935-941.
142. Straub RK, Barrack AJ, Cannon J, Powers CM. Trunk Inclination During Squatting is a
Better Predictor of the Knee-Extensor Moment Than Shank Inclination. J Sport Rehabil.
2021:1-6.
143. Straub RK, Horgan A, Powers CM. Clinical Estimation of the Use of the Hip and Knee
Extensors During Athletic Movements Using 2D Video. J Appl Biomech.
2021;37(5):458-462.
144. Straub RK, Horgan A, Powers CM. Estimation of vertical ground reaction force
parameters during athletic tasks using 2D video. Gait Posture. 2021;90:483-488.
145. Straub RK, Powers CM. Utility of 2D Video Analysis for Assessing Frontal Plane Trunk
and Pelvis Motion during Stepping, Landing, and Change in Direction Tasks: A Validity
Study. Int J Sports Phys Ther. 2022;17(2):139-147.
146. Tarara DT, Hegedus EJ, Taylor JB. Real-time test-retest and interrater reliability of select
physical performance measures in physically active college-aged students. Int J Sports
Phys Ther. 2014;9(7):874-887.
147. Taylor JB, Wright AA, Dischiavi SL, Townsend MA, Marmon AR. Activity Demands
During Multi-Directional Team Sports: A Systematic Review. Sports Med.
2017;47(12):2533-2551.

106
148. Taylor JM. Choosing the number of controls in a matched case-control study, some
sample size, power and efficiency considerations. Stat Med. 1986;5(1):29-36.
149. Teng HL, Powers CM. Sagittal plane trunk posture influences patellofemoral joint stress
during running. J Orthop Sports Phys Ther. 2014;44(10):785-792.
150. Tomescu SS, Bakker R, Beach TAC, Chandrashekar N. The Effects of Filter Cutoff
Frequency on Musculoskeletal Simulations of High-Impact Movements. J Appl Biomech.
2018;34(4):336-341.
151. Tsai LC, Ko YA, Hammond KE, et al. Increasing hip and knee flexion during a drop-
jump task reduces tibiofemoral shear and compressive forces: implications for ACL
injury prevention training. J Sports Sci. 2017;35(24):2405-2411.
152. Tsai LC, Powers CM. Increased hip and knee flexion during landing decreases
tibiofemoral compressive forces in women who have undergone anterior cruciate
ligament reconstruction. Am J Sports Med. 2013;41(2):423-429.
153. Webster KE, Hewett TE. What is the Evidence for and Validity of Return-to-Sport
Testing after Anterior Cruciate Ligament Reconstruction Surgery? A Systematic Review
and Meta-Analysis. Sports Med. 2019;49(6):917-929.
154. Weinhandl JT, O'Connor KM. Assessment of a greater trochanter-based method of
locating the hip joint center. J Biomech. 2010;43(13):2633-2636.
155. Weir G, Alderson J, Smailes N, Elliott B, Donnelly C. A Reliable Video-based ACL
Injury Screening Tool for Female Team Sport Athletes. Int J Sports Med.
2019;40(3):191-199.
156. Wiggins AJ, Grandhi RK, Schneider DK, et al. Risk of Secondary Injury in Younger
Athletes After Anterior Cruciate Ligament Reconstruction: A Systematic Review and
Meta-analysis. Am J Sports Med. 2016;44(7):1861-1876.
157. Willson JD, Davis IS. Lower extremity mechanics of females with and without
patellofemoral pain across activities with progressively greater task demands. Clin
Biomech (Bristol, Avon). 2008;23(2):203-211.
158. Willson JD, Davis IS. Utility of the frontal plane projection angle in females with
patellofemoral pain. J Orthop Sports Phys Ther. 2008;38(10):606-615.

107
159. Willson JD, Ireland ML, Davis I. Core strength and lower extremity alignment during
single leg squats. Med Sci Sports Exerc. 2006;38(5):945-952.
160. Wright RW, Dunn WR, Amendola A, et al. Risk of tearing the intact anterior cruciate
ligament in the contralateral knee and rupturing the anterior cruciate ligament graft
during the first 2 years after anterior cruciate ligament reconstruction: a prospective
MOON cohort study. Am J Sports Med. 2007;35(7):1131-1134.
161. Wu J. Cluster analysis and K-means clustering: an introduction. In: Advances in K-means
Clustering. Springer; 2012:1-16.
162. Zadpoor AA, Nikooyan AA. The relationship between lower-extremity stress fractures
and the ground reaction force: a systematic review. Clin Biomech (Bristol, Avon).
2011;26(1):23-28.

APPENDIX:
SAMPLE SURVEY

108
Q1 Information Sheet

Title: ACL Return to Sport Movement Evaluation: A Prospective Study

Principal Investigator: Christopher M. Powers, PT, PhD

Department: USC Division of Biokinesiology and Physical Therapy

You are receiving this survey because you previously underwent return to sport testing
at the Movement Performance Institute (MPI). The purpose of this survey is to help
us determine if your Return to Sport score predicts future re-injury. We do not
anticipate that taking this survey will contain any risk or inconvenience to you.
Furthermore, your participation is strictly voluntary, and you may withdraw your
participation at any time without penalty. Your participation will help us understand
and perhaps minimize the chance of re-injury for people who have ACL
reconstruction.

We will be questioning you about the period following your 2019 MPI evaluation ONLY.
Your 2019 evaluation is either (a) your most recent test (and we have no records on your
thereafter) or (b) your last test prior to your 1st subsequent ACL injury (which we have
records of, but we wish to verify and obtain additional information).

You will not receive compensation for taking part in this survey.

All information collected will be used only for our research and will be kept
confidential. There will be no connection to you specifically in the results or in future
publication of the results. Once the study is completed, we would be happy to share
the results with you if you desire. In the meantime, if you have any questions, please
ask or contact Dr. Christopher Powers (powers@usc.edu or 323.442.1928).

If you have questions, concerns, or complaints about the research and are unable to
contact the research team, contact the Institutional Review Board (IRB) Office at
323-223-2340 between the hours of 8:00 AM and 4:00 PM, Monday to Friday. (Fax:
323-224-8389 or email at irb@usc.edu).

If you have any questions about your rights as a research participant, or want to talk
to someone independent of the research team, you may contact the Institutional
Review Board Office at the numbers above or write to the Health Sciences
Institutional Review Board at LAC+USC Medical Center, General Hospital Suite
4700, 1200 North State Street, Los Angeles, CA 90033.

By selecting YES you are verifying that you have read the explanation of the study
and that you agree to participate. You also understand that your participation in this
study is strictly voluntary.
o Yes
o No

Skip To: Q2 If Q1 = Yes
Skip To: End of Survey If Q1 = No

109

Q2 Did you ever return to your sport or activity following your ACL reconstruction
(specifically after your 2019 MPI evaluation)?
o Yes
o No

Q3 At your 2019 MPI evaluation, was this your 1st tear? If not, how many previous
tears did you have at that time?
o Yes
o No ________________________________________________

Q4 What is your gender?
o Male
o Female
o Other ________________________________________________

Display This Question:
If Q2 = Yes

Q5 How old were you in years when you returned to your sport or activity
(specifically after your 2019 MPI evaluation)?
________________________________________________________________

Display This Question:
If Q2 = No

110
Q6 Why didn't you return to your sport or activity? Please select all that apply.
▢ Fear of re-injury
▢ Pain
▢ Weakness
▢ Loss of interest
▢ Lack of confidence
▢ Other (please explain)
________________________________________________

Display This Question:
If Q2 = No

Q7 Have you modified your life style to avoid potentially damaging activities to your
knee?
o Not at all
o Mildly
o Moderately
o Severely
o Totally

Display This Question:
If Q2 = No

111
Q8 In general, how much difficulty do you have with your knee?
o None
o Mild
o Moderate
o Severe
o Extreme

Display This Question:
If Q2 = Yes

Q9 Since returning to your sport or activity, have you re-injured your ACL (or torn
your ACL on the opposite side) (specifically after your 2019 MPI evaluation)?
o Yes
o No

Display This Question:
If Q9 = Yes
And Q2 = Yes

Q10 Was your 1st subsequent ACL tear (which took place after your 2019 MPI
evaluation) on the same side as your previous tear?
o Yes
o No

Display This Question:
If Q2 = Yes
And Q9 = Yes

112
Q11 Did your 1st subsequent ACL tear (which took place after your 2019 MPI
evaluation) occur while participating in your sport or activity? If yes, which sport or
activity?
o Yes ________________________________________________
o No

Display This Question:
If Q2 = Yes

Q12 Which leg is your dominant leg (i.e., one you prefer to kick a ball with)?
o Right
o Left
o No preference

Display This Question:
If Q2 = Yes

Q13 Who cleared you to return to your sport or activity (specifically after your 2019
MPI evaluation)? Check all that apply.
▢ Physician
▢ Physical therapist
▢ Coach
▢ Athletic trainer
▢ I was not cleared
▢ Other (please explain)
________________________________________________

113
Display This Question:
If Q2 = Yes

Q14 What year did you return to your sport or activity (specifically after your 2019
MPI evaluation)?
o 2019
o 2020
o 2021
o I don't remember

Display This Question:
If Q2 = Yes

114
Q15 What month did you return to your sport or activity (specifically after your
2019 MPI evaluation)?
o Jan
o Feb
o March
o April
o May
o June
o July
o Aug
o Sept
o Oct
o Nov
o Dec
o I don't remember

Display This Question:
If Q2 = Yes

115
Q16 What sport or activity did you return to immediately upon release (specifically
after your 2019 MPI evaluation)? Select all that apply.
▢ Baseball
▢ Basketball
▢ Boxing
▢ Cycling
▢ Dance
▢ Field Hockey
▢ Football
▢ Golf
▢ Gymnastics
▢ Handball
▢ Ice Hockey
▢ Ice Skating
▢ Karate
▢ Lacrosse
▢ Rugby
▢ Running or Cross Country

116
▢ Skiing
▢ Snowboarding
▢ Soccer
▢ Swimming
▢ Tennis
▢ Track & Field
▢ Volleyball
▢ Wrestling
▢ Other (please explain)
________________________________________________

Display This Question:
If Q2 = Yes

117
Q17 How often did you perform the following activities immediately upon release
(specifically after your 2019 MPI evaluation)?

Less than 1
time a
month
1 time in a
month
1 time in a
week
2 or 3 times
in a week
4+ times in
a week
Running
o o o o o
Cutting
o o o o o
Deceleration
(i.e., coming
to a quick
stop while
running)
o o o o o
Pivoting (i.e.,
turning your
body with
your foot
planted while
playing your
sport or
activity)
o o o o o

Display This Question:
If Q2 = Yes

Q18 Once your returned to your sport or activity, did you participate in a pivoting
and/or cutting sport for at least 50 hours during the following times?
Yes No NA
Year 1
o o o
Year 2
o o o
Year 3
o o o

Display This Question:
If Q2 = Yes

118
Q19 Once you returned to your sport or activity, did you participate in a formal ACL
injury prevention program (specifically during the 1st 12 months after your 2019 MPI
evaluation)? If yes, how many days a week?
o Yes ________________________________________________
o No

Q20 Did you continue your exercise program prescribed to you by your health care
provider at the time of release for sport or activity (specifically during the 1st 12
months after your 2019 MPI evaluation)? If yes, how many days a week?
o Yes ________________________________________________
o No

Display This Question:
If Q2 = Yes

Q21 What is the highest level of sport or activity you returned to immediately upon
release (specifically after your 2019 MPI evaluation)?
o Middle School
o High School
o College
o Professional
o Recreational
o Other (please explain)
________________________________________________

Display This Question:
If Q2 = Yes

Q22 What level of sport or activity did you return to immediately upon release
(specifically after your 2019 MPI evaluation)? Enter a percentage (i.e., 100%, 80% of
previous level, etc.).
________________________________________________________________

119

Display This Question:
If Q9 = Yes
And Q2 = Yes

Q23 How did your 1st subsequent ACL injury occur (specifically after your 2019
MPI evaluation)?
o Another player hit me (but not my knee)
o Another player hit my knee
o No one hit me
o I don't remember
o Other (please explain)
________________________________________________

Display This Question:
If Q9 = Yes
And Q2 = Yes

Q24 What year was your 1st subsequent ACL injury (specifically after your 2019 MPI
evaluation)?
o 2019
o 2020
o 2021
o I don't remember

Display This Question:
If Q2 = Yes
And Q9 = Yes

120
Q25 What month was your 1st subsequent ACL injury (specifically after your 2019
MPI evaluation)?
o Jan
o Feb
o March
o April
o May
o June
o July
o Aug
o Sept
o Oct
o Nov
o Dec
o I don't remember

Q26 May we contact you if we need clarification (or additional information) on
anything? If yes, please provide your email.
o Yes ________________________________________________
o No

Q27 If there are any details you would like to add regarding any of your answers (or
anything else you think we should know), please do so here.
________________________________________________________________

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Whole body mechanics of running turn maneuvers: relationship to injury and performance

PDF

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

PDF

Altered gait mechanics reduce sagittal plane knee loading following ACL reconstruction

PDF

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

PDF

Hip and pelvis kinematics and kinetics in persons with femoroacetabular impingement

PDF

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

PDF

Dynamic knee loading asymmetries following anterior cruciate ligament reconstruction: methods for clinical detection

PDF

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

PDF

Control and dynamics of turning tasks with different rotation and translation requirements

Asset Metadata

Creator Straub, Rachel Kathleen (author)

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

School School of Dentistry

Degree Doctor of Philosophy

Degree Program Biokinesiology

Degree Conferral Date 2022-08

Publication Date 06/13/2022

Defense Date 05/26/2022

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

Tag biomechanics,knee kinematics,knee kinetics,movement Screening,OAI-PMH Harvest,two-dimensional video analysis

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Powers, Christopher (committee chair), Keim, Bob (committee member), Michener, Lori (committee member), Salem, George (committee member), Sigward, Susan (committee member)

Creator Email rstraub@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC111339685

Unique identifier UC111339685

Document Type Dissertation

Format application/pdf (imt)

Rights Straub, Rachel Kathleen

Internet Media Type application/pdf

Type texts

Source 20220613-usctheses-batch-946 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

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

Repository Email cisadmin@lib.usc.edu