Investigation of Intrinsic Spine Muscle Properties to Improve Musculoskeletal Spine Modelling

Home , Erector spinae muscles, Intertransversarii, Intertransverse ligament, Latissimus dorsi muscle, Multifidus muscle

Investigation of intrinsic spine muscle properties to improve musculoskeletal spine modelling

Derek Peter Zwambag

A Thesis Presented to The University of Guelph

In partial fulfillment of requirements for the degree of Doctor of Philosophy In Human Health and Nutritional Sciences

Guelph, Ontario, Canada ©Derek Zwambag, October 2016

ABSTRACT

Investigation of intrinsic spine muscle properties to improve musculoskeletal spine modelling

Derek Peter Zwambag Advisor: University of Guelph, 2016 Dr. Stephen H.M. Brown

Spine muscles are known to generate large compressive loads and play a vital role in spine stabilization. Spine loads and stability are often estimated using computational models; yet, models cannot account for inherent differences in intrinsic muscle properties, as these data are unavailable. This dissertation was borne out of this need to further understand the characteristics of spine muscles.

Part A of this dissertation consisted of three experiments each designed to address a specific research question. Each experiment also generated normative data, which were combined in Part B to create a custom musculoskeletal spine model capable of predicting dynamic active and passive muscle moments.

Generic muscle models do not accurately predict whole muscle passive stresses. Experiment I investigated passive muscle stress differences following facet joint injury. Passive muscle stresses were not altered 28 days following injury. Data from control animals were used to model passive muscle stresses throughout physiological sarcomere lengths.

Experiment II was designed to determine the sarcomere lengths of spine muscles based on posture. Physiological sarcomere lengths were measured from human cadavers in a neutral posture using laser diffraction; sarcomeres of muscles posterior to the spine were shorter than muscles anterior to the spine. During modelled flexion, posterior muscles became strained and sarcomeres approached

the optimal length for active force production. These data were incorporated into the model to estimate sarcomere lengths of spine muscles based on spine posture.

Spine muscles display a unique phenomenon known as ‘flexion relaxation’, which occurs when extensor muscles ‘turn off’ despite substantial demand near full trunk flexion. Passive tissues are believed to support the weight of the upper body.

Experiment III further tested this proposed mechanism using a pulley system to manipulate the weight of the upper body. These data were used to validate the active and passive muscle moments predicted by the musculoskeletal model.

The model predicted: a) the occurrence of flexion relaxation; b) that decreasing the external moment caused flexion relaxation to occur earlier; and c) the requirement for abdominal muscle activity at full flexion. The model also suggested that muscle generates greater passive moments than ligaments in full flexion.

iii

Acknowledgements

Thank you to my examining committee members Drs. Jack Callaghan, Geoff Power, John Srbely, and Stephen Brown. I appreciate the time you put into reading this thesis and all your comments and suggestions. Your efforts and passion for biomechanics have improved the quality of this thesis. I would also like to thank my comprehensive examination committee members Drs. Howard Vernon, Cheryle Séguin, Karen Gordon, and Stephen Brown. I was fortunate to be able to study the spine from the points of view of a clinician, biologist, engineer, and biomechanist. I am a better researcher thanks to all of you!

I can’t thank my advisor Dr. Stephen Brown enough. Thank you very much for allowing me to study a broad range of interests. I am grateful for all your guidance, motivation, and scientific insight. You are an amazing role model, which I will strive to emulate moving forward.

Thanks to Shawn Beaudette, my fellow Brown lab PhD student. You’ve been a great friend, teammate, housemate, labmate, and co-author. I’m excited to see all that you will accomplish in the future. A big thank you to Kelsey Gsell, Dennis Larson, Lydia Frost, Alex Harriss, Grace Glofcheskie, and the extended biomechanics group and faculty for making HHNS a great place to work and learn.

Finally, thank you very much to my parents Gerald and Anja Zwambag, my fiancée Jacqueline Nixon, and my siblings for supporting me and encouraging me throughout this very long process. Thanks for allowing me to pursue my dream and stay in school for this long.

Table of Contents

ABSTRACT…………………………………………………………………………… ii Acknowledgements…………………………………………………………….. iv Table of contents………………………………………………………………... v List of tables……………………………………………………………………….. viii List of figures……………………………………………………………………… ix

Chapter 1: Introduction………………………………………………………. 1 1.1 Passive muscle modelling…………………………………………………… 1 1.2 Muscle anatomy………………………………………………………………….. 5 1.2.1 Multifidus………………………………………………………………….. 6 1.2.2 Erector spinae……………………………………………………………. 7 1.2.3 Latissimus dorsi…………………………………………………………. 8 1.2.4 Quadratus lumborum…………………………………………………. 8 1.2.5 Psoas major…………………..…………………………………………… 8 1.2.6 Abdominal muscles…………….……………………………………… 9 1.3 Dissertation overview……………….………………………………………… 10 1.3.1 Part A: Experiment 1………….……………………………………… 10 1.3.2 Part A: Experiment 2………….……………………………………… 11 1.3.3 Part A: Experiment 3………….……………………………………… 11 1.3.4 Part B: Musculoskeletal Model…………………………………… 11 1.4 Statement of ethics……………….…………………………………………..… 12

Chapter 2: Part A- Experiment I………………………………………….. 13 Disruption of erector spinae aponeurosis and facet compression in a rat model does not alter the passive mechanics of spine muscles after four weeks of recovery 2.1 Chapter overview……………….………………………………….....………… 13 2.2 Introduction……………….………………………………………………….…… 13 2.3 Methods……………….……………………………………………………………… 16 2.3.1 Surgical groups………………………………………………………….. 16 2.3.2 Tissue samples………………………………………………………..…. 18 2.3.3 Mechanical testing……………………..………………………………. 19 2.3.4 Data analysis……………………………..………………………………. 20 2.3.5 Behavioural testing……………………...……………………………. 21 2.3.6 Statistical analysis……………………………..………………………. 22 2.4 Results……………….……………………………………………………………..… 22 2.4.1 Elastic moduli……………………………………………………………. 22 2.4.2 Slack length………………………………………………………………. 24 2.4.3 Passive stress-length relationship………………………...……. 26 2.4.4 Behavioural testing……………………………………………………. 27 2.4.5 Histology……………………………………..……………………………. 27 2.5 Discussion……………….…………………………………………….…………… 28

2.5.1 Effect of Incision surgery…………………………………………… 29 2.5.2 Effect of Compression surgery ……………………….………… 30 2.5.3 Future directions………………………………………….….………. 32 2.6 Bridge summary……………….………………………….…………………… 34

Chapter 3: Part A- Experiment II…………………..…………………... 39 Sarcomere length organization as a design for cooperative function amongst all lumbar spine muscles 3.1 Chapter overview……………….………………….………………………… 39 3.2 Introduction……………….…………………………………….……………… 40 3.3 Materials and methods……………….……………….…………………… 43 3.3.1 Cadaveric donors …………………………….…..…………………. 43 3.3.2 Sarcomere length measurement………………………………. 44 3.3.3 Modelled operating ranges………………………………………. 45 3.3.4 Statistical analysis…………………………………………………... 48 3.4 Results……………….…………………………………………………………….. 48 3.4.1 Neutral spine cadaveric sarcomere lengths……………… 48 3.4.2 Modelled sarcomere lengths……………………………………. 50 3.4.2.1 Flexion and extension…………..…………………………. 50 3.4.2.2 Lateral bending………………………………………………. 54 3.4.2.3 Axial rotation…………………………………………………. 55 3.5 Discussion……………….……………………………………………………..… 55 3.6 Bridge summary……………….……………………………………………… 61

Chapter 4: Part A- Experiment III……………………………,,………. 63 Decreasing the required lumbar extensor moment induces earlier onset of flexion relaxation 4.1 Chapter overview……………….……………………………………….…… 63 4.2 Introduction……………….………………………………………………….… 64 4.3 Methods……………….…………………………………………………………… 67 4.3.1 Participant characteristics…………..…………………………... 67 4.3.2 Experimental set-up…………..…………………………………….. 67 4.3.3 Protocol…………..………………………………………………………. 70 4.3.4 Data processing…………..……………………………………………. 71 4.3.5 Statistical analysis…………..……………………………..…………. 72 4.4 Results……………….……………………………………………………………… 73 4.4.1 Lumbar erector spinae…………….………..………………………. 73 4.4.2 Abdominal muscles…………………..…………..……….…………. 75 4.4.3 Additional muscles…………..………………………………………. 78 4.5 Discussion……………….……………………………….……………………..… 80 4.6 Bridge summary……………….…………………………….………………… 86

Chapter 5: Part B- Musculoskeletal model………………………… 88 5.1 Chapter overview……………….…………………………………………..… 88

5.2 Passive muscle model……………….……………………………………… 90

5.2.1 Model development…………..…………………………….………. 90 5.2.2 Additional passive muscle models…………..……..…………. 93 5.2.3 Results…………..…………………………………………………..……. 95 5.2.3.1 Passive muscle stress-length relationship………… 95 5.2.3.2 L4-L5 sagittal passive muscle moment……………... 96 5.2.4 Implications of these findings…………..…………………….…. 99 5.3 Model of flexion relaxation……………….…………………………..…… 103 5.3.1 Model development…………..……………………………………... 103 5.3.2 Results of the model…………..……………………..………………. 108 5.3.2.1 Moments throughout trunk flexion movements…. 108 5.3.2.2 Moments at the critical point of flexion relaxation 110 5.3.3. Implications of these findings…………..………………………. 111

Chapter 6: Conclusions and future directions……………..…..… 114

Chapter 7: References…………………………………………………..…… 116

vii

List of tables

Table 2.1: Effects of surgical group, muscle, and sample type on the passive elastic moduli of rat spine muscles. 23

Table 2.2: Effects of surgical group, muscle, and muscle size on the slack sarcomere length of rat spine muscles. 25

Table 2.3: Spline coefficients given in pp-form that specify the passive stress-length relationships of single fibres and bundles of fibres from spine muscles of control rats. 37

Table 3.1: Location and attachment sites of all muscles fascicles that were sampled. Samples were taken in the approximate mid-belly of the muscle between attachment sites. 44

Table 3.2: Modeled intervertebral rotations, taken as population averages (White and Panjabi, 1990; McGill 2007). 46

Table 3.3: Sarcomere lengths (μm) of all muscles measured from an approximate neutral spine position in cadaveric donors. 49

Table 3.4: Modeled sarcomere lengths (µm) and percent change (below in brackets) from the neutral posture for each muscle of interest at end ranges of lumbar spine motion. 51

Table 5.1: Contribution of ligament, passive muscle, and active muscle tissue to the required L4-L5 extensor moment at the critical point of flexion relaxation. 111

viii

List of figures

Figure 1.1: (A) Sagittal view and (B) posterior views of the lumbar spine and sacrum. Select muscle fascicles of multifidus are shown originating from the L2 Spinous process and inserting on the mammillary processes of L5 and the sacrum. Abbreviations: IC, iliac crest; IVD, intervertebral disc; L1-L5, lumbar vertebrae; SP, spinous process; TP, transverse process. llustrations are reprinted from MacIntosh JE, Valencia F, Bodguk N, and Munro RR. “The morphology of the human lumbar multifidus” Clinical Biomechanics (1986) vol 1(4): 196-204 with permission from Elsevier. 2

Figure 1.2: Rabbit tibialis anterior active (open triangles) and passive (closed triangles) forces recorded isometrically at various muscle lengths. Modeled active (solid line) and passive (dashed line) muscle forces can also be seen. Figure recreated from Winters TM, Takahashi M, Lieber RL, and Ward SR “Whole muscle length-tension relationships are accurately modeled as scaled sarcomeres in rabbit hindlimb muscles” Journal of Biomechanics (2011) vol 44(1): 109-115 with permission from Elsevier. 3

Figure 1.3: Musculature of the lumbar spine. (A) Deep and (B) superficial posterior views. (C) Deep and (D) superficial anterior views. Images obtained with permission from Essential Anatomy 5 by 3D4Medical Apps. 6

Figure 2.1: Images of the dorsum of a rat undergoing facet compression surgery (A- C). (A) Incision through skin and thoracolumbar fascia. (B) Incision through erector spinae aponeurosis and blunt dissection of the fascia between the lumbar erector spinae and the multifidus muscle. The tendon of the multifidus muscle can be seen clearly in the centre of the incision. (C) Exposure of the L5-L6 facet capsule. (D) Dissection 28 days following surgery. The sutures in the erector spinae aponeurosis can still be seen. 18

Figure 2.2: Illustration of average stress compared to average quadratic fit from two theoretical passive muscle tests. The slack sarcomere lengths of muscle fibres #1 and #2 are 1.90 and 2.65 µm, respectively. The dashed lines represent the quadratic fits. Note that the black average stress and average quadratic lines diverge at sarcomere lengths less than 2.65 µm. The yellow area highlights where the average quadratic fit underestimates the average stress when some fibres are slack. 21

Figure 2.3: Passive elastic moduli of single fibres and bundles of muscle fibres at a sarcomere length of 3.2 µm. There were no differences between surgical groups. Letters indicate statistical differences between muscles and symbols indicate statistical differences between sample type (α = 0.05). 23

Figure 2.4: Box and whisker plot of the passive elastic moduli of bundles of multifidus fibres. Box indicates median, 25th and 75th percentiles. Whiskers extend

to minimum and maximum values excluding outliers, which are shown individually (*). 24

Figure 2.5: Slack sarcomere length of singles fibres and bundles of muscle fibres There were no differences between surgical groups. Letters indicate statistical differences between muscles and symbols indicate statistical differences between sample type (α = 0.05). 25

Figure 2.6: Passive stress-length relationship of single fibres and bundles of fibres of multifidus, lumbar erector spinae, and thoracic erector spinae. Thin lines are mean ± 1 sem. 26

Figure 2.7: Passive stress-length relationship of spine muscles. Larger stresses developed within multifidus single fibres and bundles of fibres at longer sarcomere lengths compared to lumbar and thoracic erector spinae. 27

Figure 2.8: Axial slices through the L5-L6 facet joints stained with Safranin-O. Left side of each image is the side assigned to either compression (A & C) or sham surgeries (B & D); right side of each image is the contralateral facet joint, which is used as a within-animal control. (A) and (B) are from animals assigned to the Compression and Incision surgeries, respectively; no histological differences were observed between Compression and Incision groups. (C) and (D) are reprinted from Henry JL, Yashpal KY, Vernon H, Kim J, and Im HJ (2012) Lumbar facet joint compressive injury induces lasting changes in local structure, nociceptive scores, and inflammatory mediators in novel rat model. Note the red staining of the compressed facet in the current study (A) compared to the study by Henry et al., (C, 2012) where obvious facet cartilage can be observed. 28

Figure 2.9: (A) Lateral view, (B) Caudal view, and (C) dorsal view of the L4-L6 motion segments of the rat spine. The left L5-L6 facet joints are circled in (A) and (C). This spine was one of the cadavers used for pilot testing of the Compression surgery. Bony damage of the facet joint can be seen in (C). 32

Figure 2.10: Comparison of passive elastic moduli values of spine muscles. Results from mouse, rat, and rabbit muscle all come from healthy tissue, while human muscle is obtained from patients receiving lumbar interbody fusion. Standard error values are not present for Ward et al., 2009b as these data were published with respect to strain (δσ/δɛ) rather than sarcomere length (δσ/δSL). Moduli values with respect to sarcomere length were calculated with ɛ = (sarcomere length – slack length)/slack length. 35

reported in the literature. Note that Ward and colleagues did not attempt to predict passive muscle stresses; these data simply illustrate that slack lengths and elastic moduli are not sufficient for accurate passive muscle stress predictions. 37

Figure 3.1: Visual representation of the computational lumbar spine model (Cholewicki & McGill 1996) with joint centres of rotation and muscle lines of action in (A) neutral position in frontal plane, (B) lateral bend in frontal plane, (C) neutral position in sagittal plane, and (D) flexion in sagittal plane. Extension and axial twist are not shown. Axes to scale in centimeters. (Abbreviations: ES, Erector spinae; IT, Intertransversarii; L1-L5, lumbar vertebrae; M L1 and M L4, Multifidus to L1 and L4; PM L1 and L4, Psoas major to L1 and L4; QL, Quadratus lumborum; S1, first sacral vertebra). 47

Figure 3.2: Sarcomere length operating ranges (μm) of muscles attaching to L1 (except Intertransversarii, which attaches between the L4 and L5 vertebrae). X = the measured sarcomere length in a neutral spine posture. (A) ▲: the modeled sarcomere length in full extension and △: modeled sarcomere length in full flexion. The box depicts how all sarcomeres of all muscles travel to the same region on the descending limb as the lumbar spine reaches full flexion. (B) Sarcomere operating ranges in lateral bend. ▲: ipsilateral bend; △: contralateral bend. (C) Sarcomere operating ranges in axial rotation. ▲: ipsilateral rotation; △: contralateral rotation. (Abbreviations: IT, Intertransversarii; QL, Quadratus lumborum; Mult, Multifidus; Long, Longissimus lumborum; Ilio, Iliocostalis lumborum; P.Mn, Psoas major; P.Mn, Psoas minor). 53-54

Figure 3.3: Transverse cross-section of the lumbar spine showing the lumbar vertebrae and paraspinal muscles. (A) Sarcomere lengths (µm) measured from lumbar spine muscles in a neutral cadaveric position and (B) modelled sarcomere lengths at full spine flexion. Blue, white, and red corresponds to sarcomeres on the ascending limb, plateau, and descending limb of the sarcomere force length curve, respectively. M, multifidus; L, longissimus; I, iliocostalis; QL, quadratus lumborum; IT, intertransversarii; and P, psoas major. 56

Figure 4.1: Schematic of experimental set up. Forces and moments were recorded from an AMTI forceplate (2048 Hz) while kinematic data were recorded from T12, S1, greater trochanter, lateral knee, lateral malleolus, and head of 5th metatarsal (32 Hz). Lumbar flexion angle (Φ) was defined as the relative angle between T12 and S1 rigid bodies. Trunk inclination angle (Ψ) was defined as the absolute angle of the T12 rigid body with respect to horizontal. A fixed pulley attached at ~T5 was used to generate an extension moment about L4-L5. 70

There was no effect of condition on trunk inclination. Different letters indicate post- hoc statistically different means between conditions. 73

Figure 4.3: Average lumbar flexion and trunk inclination angles, L4-L5 moments, and lumbar erector spinae activity of all time-normalized trials. Vertical lines denote time-normalize trials. During trunk flexion, lumbar erector spinae activity is reduced as greater masses are added to the pulley. Note that due to the experimental set-up the pulley only became engaged at ~20° of lumbar flexion. Note also that during trunk extension, hip extension occurs before lumbar extension, thereby causing trunk inclination to initiate prior to lumbar spine movement. 75

Figure 4.4: Difference in time between abdominal muscles turning ‘on’ and lumbar erector spinae turning ‘off’. There were no differences in abdominal and lumbar erector spinae muscles turning on/off in the 0, 2.27, and 4.54 kg conditions. Both rectus abdominus and external oblique activated before lumbar erector spinae inactivation in the 6.80 kg condition indicating that participants had a period of co- contraction in this condition. * indicates significant difference between abdominal muscle activation and lumbar erector spinae inactivation (α = 0.05). 76

Figure 4.5: L4-L5 moment, lumbar flexion and trunk inclination angles at the critical point of rectus abdominus activation. L4-L5 moment and lumbar flexion angles decreased as greater masses were attached to the pulley. There was no effect on trunk inclination at the critical point of rectus abdominus activation. Different letters indicate post-hoc statistically different means between conditions. 77

Figure 4.6: L4-L5 moment, lumbar flexion and trunk inclination angles at the critical point of external oblique activation. L4-L5 moment decreased as greater masses were attached to the pulley. There was no effect on lumbar flexion or trunk inclination at the critical point of external oblique activation. Different letters indicate post-hoc statistically different means between conditions. 77

Figure 4.7: Average time-normalized muscle activity of all participants through trunk flexion and return to stand. Note that lumbar erector spinae activity is shown in Figure 4.3. 78

Figure 4.8: The relationship between lumbar flexion angle and L4-L5 moment when adding (Howarth and Mastragostino, 2013) or subtracting (current study) mass from the torso. 82

Figure 5.1: Overview of input parameters used in the computational spine model. The model predicts active and passive muscle and ligament moments throughout dynamic trunk flexion. 90

Figure 5.2: Sagittal view of modelled muscle fascicles of six muscle groups included in the musculoskeletal model. 92

xii

Figure 5.3: Passive muscle stress-optimal length relationships for the McGill & Norman (1986), Thelen (2003) and current passive muscle models. The Dunk et al., (2004) model has the same passive muscle stress-optimal length relationship as McGill & Norman (1986). 96

Figure 5.4: Passive muscle L4-L5 sagittal plane moments as a function of lumbar flexion/extension angle. Positive values are extensor moments and negative values are flexor moments. All models predicted similar passive muscle moments in a neutral posture; however both the McGill and Norman (1986) and Thelen (2003) models predicted much higher moments near full flexion compared to the Dunk et al., (2004) and current models. Both the Dunk et al., (2004) and the current model are similar throughout the entire range of lumbar motion. 97

Figure 5.5: L4-L5 sagittal plane moments due to passive muscle throughout lumbar flexion range of motion. The black line represents the current model and the coloured lines are the passive muscle moments predicted by the Dunk et al., (2004) model. Spine flexibility ranges from high (purple) to low (yellow). 99

Figure 5.6: Sagittal view of L4 and L5 vertebral bodes. Modelled lines of action of seven different ligaments. All ligaments are mid-sagittal except the capsular and intertransverse ligaments, which exist bilaterally. 104

Figure 5.7: Predicted supraspinous ligament forces as a function of lumbar flexion angle from three theoretical individuals with different levels of lumbar flexibility using the equations developed by McGill and colleagues (McGill & Kippers 1984; McGill & Norman 1986; McGill 1988; McGill 1992; Potvin et al., 1991). Note that the forces at maximum flexion are much greater in the least flexible individual (yellow) compared to the most flexible (blue). All forces at end range of motion exceed the failure force (black dashed line) recorded for the supraspinous ligament (Myklebust et al. 1988). 105

Figure 5.8: Ligament forces of three theoretical participants with different levels of spine flexibility predicted using the current method of evaluating ligament rest lengths. The black dashed line represents the failure force for the supraspinous ligament (Myklebust et al. 1988). 107

Figure 5.9: Average passive muscle, ligament, and total passive L4-L5 sagittal plane moments for each condition throughout the trunk flexion movement. Larger passive muscle moments in the 0 kg and 2.27 kg conditions were due small increases in participant flexion angle. The passive muscle moments were less sensitive to these small differences in flexion angle than the ligament moments. 108

Figure 5.10: (A) The average active muscle moments for each experimental condition predicted as the difference between the required extensor moment and the total passive moment. The predicted active moments of each condition very

xiii

closely resembled the lumbar erector spinae activation profiles (B) recorded from Experiment III. 110

xiv

Chapter 1: Introduction

1.1 Passive muscle modelling

Musculoskeletal models are integral to studying the spine. The lumbar spine is composed of five vertebrae separated by intervertebral discs and facet joints. This structure allows for 36 degrees of freedom of rotational and translational movements, which are controlled and stabilized by a multitude of muscles and ligaments. Biomechanical spine models are often combined with inverse dynamic analyses to estimate loads exerted on various spine structures (Anderson et al.,

1985; Stokes & Gardner Morse 1995; Cholewicki & McGill 1996, Marras & Granata

1997; Daggfeldt & Thorstensen 2003; Shirazi-Adl et al., 2002, Christophy et al.,

2012). These models have led to significant improvements in the fields of ergonomics, movement control, and low-back rehabilitation.

L1 Facet joint SP L2 TP IVD L3

L4 IC

Sacrum

Figure 1.1: (A) Sagittal view and (B) posterior view of the lumbar spine and sacrum. Select muscle fascicles of multifidus are shown originating from the L2 spinous process and inserting on the mammillary processes of L5 and the sacrum. Abbreviations: IC, iliac crest; IVD, intervertebral disc; L1-L5, lumbar vertebrae; SP, spinous process; TP, transverse process. llustrations are reprinted from MacIntosh JE, Valencia F, Bodguk N, and Munro RR. “The morphology of the human lumbar multifidus” Clinical Biomechanics (1986) vol 1(4): 196-204 with permission from Elsevier.

Muscles are a very important component of this system. In the absence of muscles, the osteoligamentous lumbar spine buckles under ~90 N of compressive force (Crisco et al., 1992); this is less than the weight of the upper body. Further,

Cholewicki & McGill (1992) have predicted that up to 18,000 N of force can act on the spine during strenuous activity. Muscles are able to maintain spine stability by storing potential energy (Bergmark 1989); the stiffness and architecture of spine muscles determines the amount of stored potential energy. Muscle stiffness has both active and passive components. When activated, cross-bridges form between actin

and myosin protein filaments (Huxley & Niedergerke 1954; Huxley & Hanson 1954).

Tension is produced in the muscle as myosin head groups change conformation during the ‘power stroke’; muscle stiffness also increases in proportion to the number of force producing cross-bridges. In addition to the active component of muscle, muscles generate passive tension as they are stretched to longer lengths. At long muscle lengths, these passive forces can exceed the active forces generated by maximal activation (Figure 1.2). Therefore, it is important for muscle models to accurately account for both the active and passive contributions to muscle tension and stiffness. While active muscle mechanics of various human and animal systems have been studied extensively (Hill 1938; Gordon et al., 1966; Edman et al., 1978;

Herzog & Leonard 1997), less is known about passive muscle mechanics.

Passive muscle mechanics may play an important role in stabilizing the spine during trunk flexion. As healthy individuals flex forward, lumbar erector spinae activation increases in order to support the upper body mass. However near full flexion, the muscles suddenly ‘turn off’, termed flexion relaxation (Floyd & Silver

1951; 1955). Stresses developed within passive spine structures, such as ligaments and intervertebral discs, are often hypothesized to generate large extensor moments, allowing the muscles to cease active force production under these conditions. In addition to ligaments and discs the passive component of muscle would also be expected to generate large moments due to the large cross sectional areas and moment arms of spine extensor muscles. A musculoskeletal spine model with a more detailed understanding of passive spine muscle mechanics would be able to estimate the relative contribution of passive muscle during flexion relaxation.

Muscle models often rely on generic passive properties scaled using physiological cross sectional area and fascicle length to account for differences between muscles (Zajac 1989). However, Winters et al. (2011) have demonstrated that these generic curves do not accurately predict whole muscle passive forces in rabbits (Figure 1.2). The poor predictability of these generic models could stem from not incorporating differences in passive muscle moduli or slack muscle length between muscles. Musculoskeletal spine models are currently unable to account for these differences, as normative data on these parameters are unavailable in the literature.

1.2 Muscle anatomy

A muscle is composed of a number of specialized cell types encased within a complex network of extracellular tissues; this structure is designed to generate tensile forces in order to move the skeleton. Muscle fibres, or myocytes, are the most obvious cell type within skeletal muscle. They are long multinucleated cells, which connect to the skeleton via tendons and aponeuroses. Within the muscle fibre are thousands of myofibrils composed of sarcomeres arranged in series. The sarcomere is the smallest force-producing unit within a muscle. Muscle fibres are arranged into bundles called muscle fascicles. Fibroblasts are smaller and scattered throughout muscles. They are responsible for creating the fibrous extracellular network, which enables force transmission between muscle fibres and from muscle fibres to tendons/aponeuroses and provides structural support to the muscles. The extracellular matrix is often subdivided into epimysium, perimysium, and endomysium. Epimysium is the connective tissue layer covering the surface of the entire muscle. Perimysum and endomysium are the secondary and tertiary structures surrounding muscle fascicles and muscle fibres respectively.

Additionally, muscles include blood vessels, nerves, and satellite cells.

There are six groups of lumbar muscles that are considered in this thesis

(Figure 1.3). Following is a brief description of their anatomy largely based on the careful dissections of Nikoli Bogduk (Bogduk 1980; Macintosh et al., 1986;

Macintosh & Bogduk 1986; Bogduk 2005).

1.2.1 Multifidus

Multifidus is complex group of muscle fascicles that originate from the spinous processes of lumbar vertebrae and attach to the mammillary processes of vertebral bodies 2-5 vertebrae inferior (Figure 1.3 A). Multifidus has relatively short fascicle lengths and large physiological cross sectional areas (Ward et al., 2009a,

2009b). It makes up the bulk of the muscle mass at the L4-L5 vertebral level. It is

important to consider that multifidus could be reclassified as a collection of many smaller muscles, as fascicles span different motion segments and are innervated segmentally from the medial branches of the dorsal rami of lumbar spinal roots. The multifidus fibres have negligible tendons. More detailed information can be found in

Macintosh et al., (1986), Macintosh & Bogduk (1986) and Rosatelli et al., (2008).

1.2.2 Erector spinae

The erector spinae is a group of three muscles: iliocostalis, longissimus, and spinalis (lateral to medial; Figure 1.3 B). Spinalis is located predominantly within the thoracic spine and is not considered further in this thesis. Bogduk (1980) has further classified iliocostalis and longissimus as either thoracic or lumbar depending on the superior attachment site. Lumbar iliocostalis and longissimus muscles originate from the iliac crest and posterior superior iliac spine; they attach superiorly to the tips of the transverse processes and accessory processes, respectively. The fascicles attaching to the lumbar vertebrae have negligible tendons. The thoracic erector spinae muscles attach to the spine and pelvis via the erector spinae aponeurosis, a highly organized, strong, flat tendon with attachments to the iliac crest, sacrum, and lumbar spinous processes. Superiorly, the thoracic iliocostalis and longissimus attach to the angle of the rib and the transverse thoracic tubercle, respectively. The erector spinae are composed of long fascicles, have large physiological cross sectional areas, and large moments arms in all three orthopaedic planes. Further details on these muscles can be found in Bogduk (1980) and Bogduk

(2005).

1.2.3 Latissimus dorsi

Latissimus dorsi is a large flat muscle that inserts on the intertubercular groove of the humerus and on the spinous processes from T7-L5, the sacrum, and iliac crest via the posterior layer of the lumbodorsal fascia (Figure 1.3 B). It is generally considered a shoulder muscle as it generates moments to extend, adduct, and internally rotate the arm. However, it also generates extension, axial twist, and lateral bend moments throughout the lower thoracic and lumbar motion segments.

Further details can be found in Gerling & Brown (2013).

1.2.4 Quadratus lumborum

Quadratus lumborum is located lateral to the spine (Figure 1.3 A). It has attachments on the iliac crest, transverse processes of the lumbar vertebrae and the inferior border of the 12th rib. It predominantly generates lateral bend moments in the lumbar motion segments, however it also generates some lumbar extension.

1.2.5 Psoas major

The psoas major is located anterolateral to the spine (Figure 1.3 C). It arises from the vertebral bodies, transverse processes, and intervertebral discs from T12-

L5 and inserts on the lesser tubercle of the femur along with the iliacus muscle; iliacus does not create generate moments in lumbar motion segments and is not considered in this thesis. The psoas major is predominantly a hip flexor muscle.

Psoas major fascicles lie very close to the spine and have small lumbar moment

arms; psoas major produces small lumbar flexion moments at L4-L5 and may generate spine extension moments at superior levels depending on spine posture.

Primarily, passive and active forces in psoas major increase spine compression.

Further details can be found in Bogduk et al., (1992).

1.2.6 Abdominal muscles

The abdominal wall is composed of four muscles: rectus abdominus, external oblique, internal oblique, and transversus abdominus (Figure 1.3 C & D). The rectus abdominus originates on the inferior costal cartilage and xiphoid process and inserts on the pubic crest and pubic symphysis. The rectus abdominus muscles are separated down the midline by the linea alba. The rectus abdominus has the longest fascicles and smallest physiological cross sectional area of all the abdominal muscles. The anterolateral walls of the abdomen are composed of the external oblique, internal oblique, and transversus abdominus muscle from superficial to deep. These muscles attach to the inferior ribs and transverse processes of the lumbar vertebrae via the medial layer of the thoracolumbar fascia. Inferiorly these muscles attach to the pubis, inguinal ligament, and linea alba. The external oblique, internal oblique, and transversus abdominus have long fascicles, tendons/aponeuroses, and large physiological cross sectional areas. These are highly complex muscles that are tightly adhered together and are able to transmit forces laterally between muscle layers (Brown & McGill 2009; Brown 2011). More detail on the complex anatomy of these muscles can be found in Brown et al.,

(2010).

1.3 Dissertation overview

The goal of this thesis was to investigate intrinsic spine muscle parameters which affect active and passive force production. The findings of these research experiments were then incorporated into a custom musculoskeletal model of the spine, designed to predict active muscle moments during dynamic trunk flexion movements. The thesis is presented in two parts.

Part A consists of three fundamental research experiments; each experiment is presented in manuscript style and addresses a specific research question. In addition to the specific research question, each experiment in Part A was also designed to generate normative data in order to enhance passive spine muscle modelling; the ‘Bridge summary’ following each manuscript describes how this research contributes to the muscle model. Part B is the generation of the computational musculoskeletal model based on the results of the experiments in

Part A. First, predicted passive muscle stresses and moments are compared to three existing passive muscle models. Second, ligament, passive and active muscle moments are predicted dynamically from experimental trunk flexion movements to determine the contributions of each tissue to flexion relaxation.

1.3.1 Part A: Experiment I

Research question: Does passive muscle remodelling following disruption of the erector spinae aponeurosis or facet compression alter the passive mechanics of spine muscles?

Contribution to muscle model: The data from the control group were used to determine predict the passive muscle stresses developed within muscles as a function of sarcomere length. .

1.3.2 Part A: Experiment II

Research question: Do the sarcomere operating ranges of paraspinal muscles reflect each muscles unique function?

Contribution to muscle model: Determine the sarcomere lengths of trunk muscles measured in a physiologically relevant posture; these data aid in predicting passive and active muscle forces throughout ranges of spine motion.

1.3.3 Part A: Experiment III

Research question: Do lumbar erector spinae muscles exhibit flexion relaxation earlier when the required lumbar moment is reduced?

Contribution to muscle model: Determine the extensor moment, lumbar spine and thoracic inclination angles when the spine is supported by passive spine structures.

1.3.4 Part B: Musculoskeletal modeling

In Part B, data from Experiments I-III were incorporated into a custom computational musculoskeletal model of the spine. First, passive muscle moments throughout the average sagittal range of motion were compared to three existing muscle models. Second, moments generated by ligaments, passive muscles, and

active muscles were calculated during a dynamic trunk flexion task (Experiment III).

Active muscle moments were compared to the lumbar muscle activation patterns.

One of the objectives of this model was to predict the ‘Flexion Relaxation

Phenomenon’ to occur.

1.4 Statement of ethics

The experiments presented in this dissertation were all conducted in accordance with the ethical guidelines of the University of Guelph and were approved by the University Research Ethics Board. Experiment I was also approved by the Animal Care Committee (AUP #2460).

Chapter 2: Experiment I

Disruption of erector spinae aponeurosis and facet compression in a rat model does not alter the passive mechanics of spine muscles after four weeks of recovery

2.1 Chapter summary

Muscles are able to passively generate large forces at long lengths; however, current generic muscle models do not accurately predict these forces. This could be because variations in the elastic moduli and slack lengths between muscles are not well understood. Further, elastic moduli of bundles of muscle fibres have been shown to increase following tenotomy or injury. The purpose of this study was to characterize the passive elastic moduli, slack lengths, and passive muscle stresses between spine muscles, as well as in response to disruption of the erector spinae aponeurosis and facet compression injury. Disruption of the erector spinae aponeurosis was hypothesized to lead to muscle remodelling and greater passive elastic moduli of muscles surrounding the injury site as well as in muscles in series with the aponeurosis. Facet compression leading to low back injury was hypothesized to also increase the passive elastic moduli of the surrounding multifidus muscles. Both hypotheses were rejected as neither disruption of the erector spinae aponeurosis nor facet compression altered the slack length, elastic moduli, or passive muscle stress-length relationship after four weeks of recovery.

2.2 Introduction

As muscle forces are difficult to quantify in vivo, muscle models are often used to predict forces in order to calculate joint stiffness (Brown & McGill 2005;

Brown & Potvin 2007), stability (Gardner-Morse et al., 1995; Cholewicki & McGill

1996; Reeves & Cholewicki 2003; Brown & McGill 2005; Brown & Potvin 2005) and loading (Chaffin 1969; Shultz et al., 1982; Marras et al., 1984; Andersson et al., 1985;

Marras & Sommerich 1991; Granata & Marras 1995; Shirazi-Adl et al., 2002).

Accurate muscle architectural parameters are critical to these predictions. One difficulty of this is that muscles are able to adapt and remodel depending on mechanical loading (Glass 2005), injury (Brown et al., 2011; Fortuna et al., 2014;

Sato et al., 2014) and disease (Fridén & Lieber 2003; Shah et al., 2004; de Bruin et al., 2014; Matthiasdottir et al., 2014; Gsell 2016). Therefore it is important to determine how muscle parameters are altered by these various challenges in order to improve current musculoskeletal models.

Passive muscle stiffness, which relates the change in passive force to the change in muscle length, is an important property of muscle as it determines the amount of elastic energy that can be stored within the muscle. Passive stiffness is proportional to the cross sectional area, inversely proportional to the length of the fascicle, and scaled by the passive elastic modulus. While there is more detailed information in the literature about cross sectional areas and fascicle lengths of muscles, passive elastic modulus has only recently begun to receive attention. It is now known that passive elastic moduli of individual fibres and bundles of fibres are different between muscles (Prado 2005; Ward et al., 2009; Regev et al., 2011) and that the elastic modulus can be altered by injury (Brown et al., 2011; Sato et al.,

2014), or disease (Friden & Lieber 2003; Gsell 2016). Injury and disease caused the passive elastic modulus to increase in some muscles and decrease in others; the

reasons for these dichotomous changes are unknown. Further testing of passive elastic modulus using different injury models will hopefully begin to elucidate whether an injury or disease will cause the modulus to increase or decrease; this will help researchers begin to identify patterns and hypothesize why these changes occur. A novel rat surgical injury model has recently been developed mimicking osteoarthritis of the spine (Henry et al., 2012). Surgical compression of rat lumbar facet joints induced cartilage degeneration and low back pain; it is unknown whether facet degeneration also leads to muscle remodelling similar to intervertebral disc degeneration (Brown et al., 2011).

In addition to injury and disease, muscle remodelling also occurs following surgery, due to incision of tendinous insertions, muscle splitting, retraction, and denervation (Kawaguchi et al., 1996; 1998). Sivhonen et al., (1993) found that severe muscle remodelling following lumbar spine surgery was correlated with poorer patient outcomes, and is a potential cause of ‘postoperative failed back syndrome’. There have been numerous studies on the effect of surgery on muscle fascicle length and cross sectional area. Tenotomy of the rat achilles tendon has been shown to decrease muscle fascicle length and sarcomere number within three weeks of surgery (Baker & Hall-Craggs 1978; Baker & Hall-Craggs 1980; Józsa et al.,

1990), likely in order to maintain an optimal range of sarcomere lengths for force production (Jamali et al., 2000). Spine surgeries lead to fatty infiltration (Sihvonen

1993) and multifidus muscle atrophy at 6 months (Hyun et al., 2007) and 1 year post surgery (Fan et al., 2010), which decrease the cross sectional area and ability of muscles to generate passive and active forces. It is unknown whether muscle

remodelling alters the passive muscle modulus following surgery. Hu et al., (2015) detected increased muscle fibrosis and collagen content histologically three weeks after multifidus muscle retraction. While fibrosis and increased collagen content may indicate an increased passive muscle modulus, Smith and Barton (2014) did not find a strong correlation between these variables. Therefore, the effects of surgery on spine muscle passive elastic moduli remain to be determined.

The purpose of this study was to determine whether muscle remodelling altered the passive mechanical properties of spine muscles following spine surgery and injury. Scar tissue deposition following the disruption of the erector spinae aponeurosis and fascial connections between multifidus and erector spinae was hypothesized to increase the passive elastic moduli and passive muscle stresses of muscles adjacent to the injury (multifidus and lumbar erector spinae) as well as muscles connected in series with the aponeurosis (thoracic erector spinae). Injury of the facet joint causing low back pain was hypothesized to further increase the passive elastic modulus of the multifidus muscle in order to stabilize the injured joint.

2.3 Methods

2.3.1 Surgical groups

Twenty-two male Sprague Dawley rats were randomized into one of three groups (Control n = 6, Incision n = 8, and Compression n = 8; the Incision group served as a sham control for the Compression group). Incision and Compression groups were anesthetised using isofluorane and injected with carprofen (5 mg/kg) and a 50/50 lidocaine/maracaine mixture (2 mg/kg). Posterior midline incisions 16

were made from the spinous processes of L3 to S1 through the skin and subcutaneous tissue. Incisions (~2 cm) were made through the thoracolumbar fascia and erector spinae aponeurosis just lateral to the left multifidus muscle

(Figure 2.1A) and the fascial connections between the erector spinae and multifidus muscles were disrupted (Figure 2.1B) to expose the left L5/L6 facet joint capsule for both Incision and Compression groups (Figure 2.1C). Fascial dissections were done carefully to minimize damage to muscle fibres of erector spinae and multifidus.

Modified forceps were applied to the facet joint of animals assigned to the

Compression group to provide supra-physiological compression for 3 minutes; this has previously been shown to cause cartilage degeneration, tactile sensitivity, and inflammation 28 days after surgery (Henry et al., 2012). The forceps were also applied to the facet joint of the Incision group without clamping down (ie no force applied) for three minutes; this was done to ensure that the Incision group could serve as a sham control for the Compression group. The erector spinae aponeurosis was sutured (Figure 2.1D) and the skin was closed using staples.

Figure 2.1: (A) Incision through skin and thoracolumbar fascia. (B) Incision through erector spinae aponeurosis and blunt dissection of the fascia between the lumbar erector spinae and the multifidus muscle. The tendon of the multifidus muscle can be seen clearly in the centre of the incision. (C) Exposure of the L5-L6 facet capsule. (D) Dissection 28 days following surgery. The sutures in the erector spinae aponeurosis can still be seen.

2.3.2 Tissue samples

Twenty-eight days after surgery, animals were euthanized via carbon dioxide asphyxiation in accordance with the guidelines set by Animal Care Committee at the university. Muscle biopsies were taken from the left lumbar erector spinae and multifidus muscles at the L4-L6 spine level and thoracic erector spinae biopsies were taken at the T10-T12 spine level. Thoracic erector spinae muscle biopsies were taken at this level as these fibres attached to the spine via the incised section of the erector spinae aponeurosis. Muscle biopsies were stored at -20°C in a

physiologic storage solution, which also permeabilized the muscle fibres (Shah &

Lieber 2003).

L5-L6 motion segments (including vertebral bodies, intervertebral discs, and facet joints) were dissected for histological examination. Motion segments were decalcified in Cal-X for 24 hours, rinsed with 70% ethanol for 72 hours, transferred to formalin for 24 hours, and embedded in paraffin wax for sectioning. Coronal sections through the L5-L6 facet joints were stained using Safranin-O to detect cartilage degeneration.

2.3.3 Mechanical testing

Muscles were stored for no longer than 14 days prior to mechanical testing.

Single muscle fibres as well as bundles of ~6-10 muscle fibres ensheathed in their extracellular matrix were carefully excised and secured to a micro-level force transducer and a motor (Aurora Scientific, Newmarket, ON). Fibres or bundles were submerged in a physiological relaxing solution to ensure that fibres were in a relaxed state (ie. no force producing actin-myosin cross-bridges) (Ward et al., 2009

JBJS). Sarcomere lengths were measured using laser diffraction (Lieber 1990).

Specifically, a 5mW diode laser (wavelength = 635 nm; beam diameter ≈ 1.4 mm) was shone through the muscle samples and the resulting diffraction patterns were recorded by a 256 element linear photodiode array. A custom Labview program was used to measure the calibrated distance between the 1st order peaks. Slack sarcomere length was defined as the length where passive tension began to develop within the muscle fibre or bundle. Muscle samples were then stretched in ~0.2

µm/sarcomere increments and allowed to stress-relax for 120 s; this length of time was sufficient for muscle forces to decline to a steady state (Ward et al., 2009).

Muscle fibre and fibre bundle diameters were measured at three points along the length of the muscle using micromanipulors with a precision of 1 µm; passive forces were converted to stress using the cross-sectional area, estimated from the average diameter assuming a cylindrical shape (Linke et al., 1994; Prado et al., 2005).

2.3.4 Data Analysis

The stress-sarcomere length of each muscle test was fit with a quadratic function, restricted so that stress was non-negative and the first and second derivatives were positive. The elastic modulus was defined as the slope of the tangent to the quadratic function at a sarcomere length of 3.2 µm; this length was chosen as it occurs on the descending limb of the force-length curve where passive forces are substantial. This length is also physiologically relevant, as the lower and upper bounds of rat spine sarcomere lengths have previously been measured to be

1.8 and 3.7 µm, respectively, in neutral and flexed postures (Zwambag, Nolan &

Brown, unpublished data).

In order to generate representative passive stress-length curves, stress was evaluated in 0.1 µm increments from 1.8 to 3.7 µm using the quadratic fits of each test. Mean passive muscle stress and the standard error of the mean (SEM) were calculated for each sarcomere length. This method of averaging each muscle test was used rather than simply averaging the quadratic coefficients, as the domains of experimental data were different lengths; averaging the quadratic coefficients

assumes that each quadratic fits the data within the entire domain of 1.8-3.7 µm, this is not true as many fibres have slack sarcomere lengths longer than 1.8 µm

(Figure 2.2).

Figure 2.2: Illustration of average stress compared to average quadratic fit for two theoretical muscle fibres. The slack sarcomere lengths of muscle fibres #1 and #2 are 1.90 and 2.65 µm, respectively. The dashed lines represent the quadratic fits. Note that the black average stress and average quadratic lines diverge at sarcomere lengths less than 2.65 µm. The yellow area highlights where the average quadratic fit underestimates the average stress when some fibres are slack.

2.3.5 Behavioural testing

Rat hind foot paw sensitivity was tested one day prior to surgery and one day prior to sacrifice using the methods described by Pitcher et al., (1999). Rats with low back pain were expected to have increased foot paw sensitivity due (Henry et al., 2012). Two rats at a time were placed in custom plexi-glass cages with holes in

the floor allowing for Von Frey monofilaments to be applied to the pad of the left hind foot. Rats were habituated for 30 minutes prior to testing. Monofilaments were applied to the left foot paw in an ascending/descending pattern to determine the threshold for foot withdrawal.

2.3.6 Statistical Analysis

Differences between elastic modulus and slack sarcomere lengths were tested using 3-way analysis of variance (ANOVA). Independent factors were Group

(control, incision, and compression), Muscle (multifidus, lumbar erector spinae, and thoracic erector spinae), and Type (single fibres and bundles of fibres). If data were not normally distributed within a factor, then differences were tested using Kruskal-

Wallis one-way analysis of variance. For passive stress-length curves, stresses were compared between muscles and between groups using an ANOVA at each sarcomere length.

2.4 Results

2.4.1 Elastic moduli

There was no main effect of surgical group on passive elastic moduli

(p = 0.5318; Figure 2.3). There were main effects of sample type (p < 0.0001; bundles of fibres have greater passive elastic moduli than single fibres) and muscle

(p = 0.0001; multifidus has greater passive elastic moduli than lumbar and thoracic erector spinae). There were no 2-way or 3-way interaction effects between group, muscle, and sample type on elastic moduli (Table 2.1).

Table 2.1: Effects of surgical group, muscle, and sample type on the passive elastic moduli of rat spine muscles. Source Sum sq df Mean sq F Prob>F Group 470.1 2 235 0.63 0.5318 Muscle 6792.9 2 3396.4 9.14 0.0001 Size 24870.6 1 24870.6 66.94 <0.0001 Group*Muscle 261.2 4 65.3 0.18 0.9508 Group*Size 1439.3 2 719.6 1.94 0.1456 Muscle*Size 1185.6 2 592.8 1.6 0.2042 Group*Muscle*Size 974.5 4 243.6 0.66 0.6232 Error 140078.4 377 371.6 Total 178394.7 394

It is evident from Figure 2.3 that the passive elastic moduli of single fibres and bundles of lumbar and thoracic erector spinae muscles were not different between the surgical groups. However, the means comparison of the multifidus bundles gave the impression that passive elastic moduli may increase from Control

to Incision to Compression groups. A Kruskal-Wallis non-parametric test confirmed that there was no effect of surgical group on multifidus passive elastic moduli

(p = 0.6755); outliers in the data set caused the apparent increase in mean passive elastic moduli (Figure 2.4).

Figure 2.4: Box and whisker plot of the passive elastic moduli of bundles of multifidus fibres. Box indicates median, 25th and 75th percentiles. Whiskers extend to minimum and maximum values excluding outliers, which are shown individually (*).

2.4.2 Slack length

There was not a main effect of surgical group on the slack sarcomere length

(p = 0.8312; Figure 2.5). There were main effects of sample type (p < 0.0001; bundles have shorter slack sarcomere lengths than single fibres) and muscle (p =

0.0021; multifidus have shorter slack sarcomere lengths than lumbar or thoracic

erector spinae). There were no 2-way or 3-way interaction effects between group, muscle, and muscle size on slack sarcomere length (Table 2.2).

Table 2.2: Effects of surgical group, muscle, and muscle size on the slack sarcomere length of rat spine muscles. Source Sum sq df Mean sq F Prob>F Group 0.0155 2 0.00777 0.18 0.8312 Muscle 0.529 2 0.26452 6.3 0.0021 Type 1.3802 1 1.38017 32.85 <0.0001 Group*Muscle 0.3106 4 0.07765 1.85 0.1191 Group*Type 0.0181 2 0.00907 0.22 0.806 Muscle*Type 0.0068 2 0.0034 0.08 0.9224 Group*Muscle*Type 0.0679 4 0.01697 0.4 0.8058 Error 15.0841 359 0.04202 Total 17.3621 376

Figure 2.5: Slack sarcomere length of singles fibres and bundles of muscle fibres There were no differences between groups. Letters indicate statistical differences between muscles and symbols indicate statistical differences between sample type (α = 0.05).

2.4.3 Passive stress-length relationship

The passive stress-length relationships of single fibres and bundles of fibres were not different between controls and surgical groups at any sarcomere length between 1.8 and 3.7 µm (Figure 2.6). Multifidus single fibres and bundles of fibres had greater passive stress than lumbar and thoracic erector spinae, especially at sarcomere lengths greater than 3.0 µm (Figure 2.7).

Figure 2.6: Passive stress-length relationship of single fibres and bundles of fibres of multifidus, lumbar erector spinae, and thoracic erector spinae. Thin lines are mean ± 1 sem.

2.4.4 Behavioural Testing

No differences in hind foot paw sensitivity were found between Incision and

Compression groups one day prior to surgery (p = 0.6200) or 27 days post surgery

(p = 0.5924). Withdrawal thresholds for Incision and Compression groups prior to surgery were 190 ± 101 and 275 ± 128 mN, and 27 days post surgery withdrawal thresholds were 70 ± 16 and 59 ± 12 mN, respectively. The large variability is due to the logarithmic scale of the Von Frey filaments.

2.4.5 Histology

There was no histological evidence of cartilage degeneration in either the

Incision or Compression groups (Figure 2.8).

2.5 Discussion

The hypotheses that the Incision group and the Compression group would have altered passive muscle properties compared to the Control group were rejected. There was no evidence that passive elastic moduli, slack sarcomere length, or the passive stress-length relationship were affected, at 28 days post surgery, by either disruption of the erector spinae aponeurosis or compression of the facet joint.

2.5.1 Effect of Incision surgery

While we hypothesized that a surgical incision through the thoracolumbar fascia and erector spinae aponeurosis would lead to passive muscle remodelling, it may be beneficial for surgical patients that this was not the case. Concerns have been raised about the level of muscular damage inflicted during spinal surgeries.

Minimally invasive surgeries have been designed for lumbar decompression, interbody fusion, and vertebral realignment with the goal of reducing iatrogenic muscle damage (Kawaguchi et al., 1996; 1998). These surgeries have been thought to be successful in reducing muscle damage compared to conventional surgeries, evidenced by reduced serum creatine kinase 3 days after surgery (Kim et al., 2006;

Fan et al, 2010) and reduced muscle atrophy at 6 months (Hyun et al., 2007) and 1 year (Fan et al., 2010). However, the first step of any spine surgery is to cut through the erector spinae aponeurosis; therefore, it is important to investigate whether this alone would stimulate rapid muscle remodelling.

Previously, surgical tenotomy of the rat achilles tendon increased the

%volume of intramuscular connective tissue from 2.8% to 38.7% in the soleus and from 3.1% to 24.4% in the gastrocnemius three weeks after surgery (Józsa et al.,

1990). Pure muscle retraction of the rabbit multifidus muscle also increased intramuscular connective tissue three weeks later quantified histologically (Hu et al., 2015). It is important to note that both the rat achilles tentomy and rabbit multifidus retraction studies did not measure passive muscle mechanics, but rather the quantity of connective tissue histologically. While the intramuscular connective

tissue is believed to affect the passive muscle mechanical behaviour, simply increasing the amount of connective tissue does not lead to subsequent increases in passive muscle stiffness. This is likely because the mechanical behaviour is more dependent on the connective tissue organization (eg orientation and cross-linking) than the quantity (Gillies & Lieber 2011).

Another important consideration is that only a partial tenotomy was performed on the erector spinae aponeurosis in the current study. Due to this, forces generated by the thoracic erector spinae could no longer be transferred directly to the L5-L6 spinous processes; it is likely that these forces were transferred instead through the intact cranial and caudal portions of the erector spinae aponeurosis. The biaxial transmission of forces through the erector spinae aponeurosis is unknown; future studies to quantify the mechanics of this tissue are warranted. Finally, it is also important to note that despite no changes in the passive stress and elastic modulus between groups, these measures do not account for changes in the structure of the muscles. Changes in either the fascicle length or cross sectional area could alter the passive forces generated by the muscle.

2.5.2 Effect of Compression surgery

The hypothesis that the elastic modulus of multifidus muscles would increase following facet compression was also rejected. However, this hypothesis was predicated on facet compression inducing cartilage degeneration and low back pain; there was no evidence of cartilage degeneration or low back pain in animals

receiving facet compression. Therefore, further testing will be needed to determine if facet injury causing low back pain alters passive muscle mechanics.

The facet compression surgery was designed from previous work by Henry et al., (2012); these authors found that 3 min of facet compression was sufficient to induce cartilage degeneration, hind paw sensitivity, pressure sensitivity, and elevated cytokine markers of inflammation in the dorsal root ganglia by 28 days post surgery. We have hypothesized three potential reasons why our facet compression model did not replicate the findings of Henry and colleges (2012).

First, we may not have applied as much pressure to the facet joints. Henry et al.,

(2012) reported that facets were compressed to ~1 mm using modified haemostats.

During our pilot testing, compression surgeries were performed on six cadaveric rats; spine muscles were removed after the surgeries to inspect the facet joint. While we successfully targeted the L5-L6 joint, dissections revealed that we were fracturing the vertebrae (Figure 2.9C). As this was not the goal of the surgery, we reduced the applied force for all compression surgeries. Second, the orientation of the haemostats had to be changed in the current study to minimize damage to the surrounding musculature. Henry and colleagues (2012) were able to place the haemostats along the rostral-caudal axis by removing the multifidus musculature.

As muscular remodelling was the primary interest of the current study, a lateral approach to the facet joint was required in order to minimize muscle damage. There is a small tubercle on the lateral border of the rostral articulating process of the caudal vertebrae. This tubercle may have acted as a wedge, reducing compression of the facet joint. The final reason we did not replicate the findings of Henry et al.,

(2012) may be because we did not damage the surrounding musculature. Removing the multifidus muscle from the damaged facet surface may have exacerbated the compression injury in the Henry et al., (2012) study, as the muscles were no longer able to stabilize and control the damaged facet joint. Further testing will need to be done in order to determine if facet compression with minimal muscle damage is sufficient to induce cartilage degeneration and low back pain.

Figure 2.9: (A) Lateral view, (B) caudal view, and (C) dorsal view of the L4-L6 motion segments of the rat spine. The left L5-L6 facet joints are circled in (A) and (C). This spine was one of the cadavers used for pilot testing of the Compression surgery. Bony damage of the facet joint can be seen in (C).

2.5.3 Future directions

While we are confident that there were no changes in passive elastic modulus and passive muscle stress four weeks following these spine surgeries,

there may be changes at later time points. Previous studies have reported changes in passive elastic modulus eight and 12 weeks after injury. Brown et al., (2011) found that puncturing two intervertebral discs led to greater passive elastic modulus of bundles of multifidus fibres after 12 weeks. Sato et al., (2014) also found that bundles of supraspinatus and infraspinatus muscles had greater passive elastic modulus eight weeks after tenotomy and chemical denervation. Four weeks was chosen for this study as Henry et al., (2012) reported that this was sufficient time to cause cartilage damage and low back pain. Both Józsa et al., (1990) and Hu et al.,

(2015) also reported that muscle fibrosis was increased by three weeks after surgery. While not statistically significant, Brown et al., (2011) also found that mean passive elastic moduli were greater four weeks after disc puncture. As the differences in elastic moduli reported by Brown et al., (2011) were lesser at four weeks than at 12 weeks, the number of animals in each surgical group were doubled in the current study in order to increase the statistical power for detecting these lesser differences.

This work indicates that passive muscle stresses and elastic moduli are not altered four weeks following incision of erector spinae aponeurosis or facet compression. This could either be because these injuries are not severe enough to require muscle remodelling or that there has not been enough time for muscle remodelling to increase passive force. Future work will need to determine the time- course of passive muscle remodelling.

2.6 Bridge summary

The results of this study were interesting largely because they were unexpected. In addition to the results of the two different surgical groups, the results of the control group also provide us with valuable insight into the passive muscle mechanics of these spine muscles. The passive elastic moduli of the bundles of multifidus were 22% greater than the lumbar erector spinae and 42% greater than the thoracic erector spinae. This resulted in greater passive stresses in the multifidus fibres and bundles at long sarcomere lengths compared to the lumbar and thoracic erector spinae. This may be an adaptation of multifidus to store a greater amount of potential energy in order to stabilize the spine. However, it is impossible to conclude whether the multifidus is ‘more important’ than the erector spinae for spine stabilization as the physiological cross sectional areas, fascicle lengths, tendon mechanics, and moment arms would all need to be considered.

While the greater multifidus moduli and passive stresses are interesting and support the popular opinion that the multifidus is primarily a spine stabilizer, there were no differences in elastic moduli between multifidus and lumbar erector spinae in mice (Zwambag et al., 2016).

Animal experiments are necessary for this research due to the ethical and medical concerns surrounding the acquisition of muscle tissue from healthy human spines. However, it is important to consider whether results of animal testing can be translated to humans. In order to determine variability between species and spine muscles, elastic moduli values were compared with values found in the literature

(Fig 2.10). Spine muscle passive moduli were between 30-50 kPa; however there is some variability between species and muscles. Humans have the most variability between muscles; this could be because tissue samples are obtained from pathological populations compared to healthy animals used for the animal studies.

The passive elastic moduli of erector spinae and multifidus of rats were also the most similar to humans; rats may be a better model for human erector spinae and multifidus muscles than either mice or rabbits.

respect to sarcomere length were calculated with ɛ = (sarcomere length – slack length)/slack length.

While the elastic modulus evaluated at a sarcomere length of 3.2 µm is useful for detecting differences between species, muscles, or experimental populations, it is not the most useful parameter for musculoskeletal modelling. Models often require passive forces to be estimated based on a given muscle length. The stress- length relationship can be used to predict passive force if the muscle cross-sectional area is known. However, many studies that investigate passive muscle mechanics only report the elastic modulus and slack length. Solely using these two pieces of information would define a linear stress-length relationship, which clearly does not match experimental data (Figure 2.11). Passive muscle models would benefit from researchers reporting the shape of passive stress-length relationships using spline coefficients (Table 2.3) rather than simply the elastic modulus and slack length. A cubic spline with three segments was found to best describe the passive stress- length relationship.

Figure 2.11: Passive muscle stresses of multifidus throughout the operating range of rat sarcomere lengths. The orange values are average stresses measured from Control rats in the current study. The purple values are estimated stresses based solely on slack length and elastic modulus of the current study. Blue values are estimated from slack lengths and elastic moduli of human multifidus muscles reported in the literature. Note that Ward and colleagues did not attempt to predict passive muscle stresses; these data simply illustrate that slack lengths and elastic moduli are not sufficient for accurate passive muscle stress predictions.

Table 2.3: Spline coefficients given in pp-form that specify the passive stress-length relationships of single fibres and bundles of fibres from spine muscles of control rats. Domain (µm) 1.800 < SL ≤ 2.433 2.433 < SL ≤ 3.067 3.067 < SL ≤ 3.700 Coefficients c1 c2 c3 c4 c1 c2 c3 c4 c1 c2 c3 c4 Bundles Multifidus 7.017 0 0 0 3.836 13.332 8.443 1.783 -5.523 20.620 29.946 13.452 LES 5.731 0 0 0 3.478 10.888 6.896 1.456 -9.209 17.497 24.873 11.074 TES 4.460 2.019 0 0 0.294 10.493 7.924 1.943 -1.036 11.053 21.570 11.245 Fibres - Multifidus 5.570 0 0 0 0.190 10.583 6.703 1.415 -3.389 10.223 19.880 9.857 LES 3.319 0 0 0 0.794 6.307 3.994 0.843 -4.113 7.815 12.938 6.104 TES 3.719 0 0 0 1.719 7.066 4.475 0.945 -5.438 10.332 15.494 7.050 *The pp-form of a spline is defined by local polynomial coefficients over a specific 3 2 domain. y = c1*(x-xA) + c2*(x-xA) + c3*(x-xA) + c4; where the domain spans [xA ,xB].

The musculoskeletal model can use these splines in order to predict passive

muscle stress at any sarcomere length. However, in order for these splines to

improve musculoskeletal spine modelling, it is important to know what the sarcomere lengths of the various muscles are, based on the posture of the spine.

Measuring the physiological sarcomere length of spine muscles in a neutral spine posture and predicting the sarcomere operating ranges of these muscles was explored in Experiment II.

Chapter 3: Experiment 2

Sarcomere length organization as a design for cooperative function amongst all lumbar spine muscles

3.1 Chapter overview

The functional design of spine muscles in part dictates their role in moving, loading, and stabilizing the lumbar spine. There have been numerous studies that have examined the isolated properties of these individual muscles. Understanding how these muscles interact and work together, necessary for the prediction of muscle function, spine loading, and stability, is lacking. The objective of this study was to measure sarcomere lengths of lumbar muscles in a neutral cadaveric position and predict the sarcomere operating ranges of these muscles throughout full ranges of spine movements. Sarcomere lengths of seven lumbar muscles in each of seven cadaveric donors were measured using laser diffraction. Using published anatomical coordinate data, superior muscle attachment sites were rotated about each intervertebral joint and the total change in muscle length was used to predict sarcomere length operating ranges. The extensor muscles had short sarcomere lengths in a neutral spine posture and there were no statistically significant differences between extensor muscles. The quadratus lumborum was the only muscle with sarcomere lengths that were optimal for active force production in a neutral spine position, and the psoas muscles had the longest lengths in this position. During modelled flexion the extensor, quadratus lumborum, and intertransversarii muscles lengthened so that all muscles operated in the approximate same location on the descending limb of the force-length relationship.

The intrinsic properties of lumbar muscles are designed to complement each other. 39

The extensor muscles are all designed to produce maximum active force in a mid- flexed posture, and all muscles are designed to operate at similar locations of the force-length relationship at full spine flexion. Further, at full spine flexion all muscles have long sarcomere lengths, where they would also be predicted to generate large passive forces.

3.2 Introduction

The spine is a highly complex structure, consisting of multiple intervertebral joints capable of multi-degree of freedom movement. Due to this complexity, its control is regulated by a vast array of seemingly uniquely designed muscles. How these muscles generate force determines spine movement, loading, and stability; thus, it is important to understand how all the muscles function and interact as a complete system. The use of electromyography (EMG) has been critical in determining how the spine muscles are activated and load the spine under a variety of conditions (e.g. Marras and Granata, 1997; Larivière et al., 2003; Kavcic et al.,

2004; Dufour et al., 2013). While EMG is highly useful for studying muscle activation, architectural and intrinsic muscle properties are required to estimate muscle force production.

The design of muscles, including their architectural and intrinsic properties, provide the basic template for force production and is one of the best predictors of muscle function (Lieber and Fridén, 2000). The sarcomere is the fundamental unit of muscle force production, and sarcomeres can be organized in many different patterns to create various physiological cross-sectional areas (PCSA) and

fibre/muscle lengths, which provide some sub-specialization for generating greater force or wider excursions, respectively. There has often been a trend of studying individual spine muscles in isolation, and then attributing isolated roles and functions to these muscles. Multifidus has been often implicated as the most important of the spinal muscles (Danneels et al., 2002; Wilke et al., 1995). However, others have argued that, while aspects of multifidus structure or function may be unique, it is essential to consider all muscles in concert, and necessary for all these muscles to work together to create an appropriately controlled and capable system

(McGill et al., 2003; Wagner et al., 2005). Interestingly, multifidus has recently been predicted to act exclusively on the ascending limb of the sarcomere force-length relationship, reaching optimal length in full spine flexion; however, in this study multifidus was the only muscle considered (Ward et al., 2009). McGill and Norman

(1986) previously predicted that all lumbar extensor muscles would share this same trait. Thus, we were motivated to assess sarcomere organization in a wider array of spine muscles.

Sarcomere length is an important intrinsic property of muscle force production as it affects binding of myosin head groups to actin filaments. The muscle active force-length relationship describes the maximal capability of forming cross-bridges at different fibre, and their constituent sarcomere, lengths (Gordon et al., 1966). Sarcomere length also influences passive force generation as intracellular titin filaments and extracellular collagen fibres become strained at long sarcomere lengths. Whole muscles each have unique active and passive force-length relationships depending on architectural parameters such as fibre length, which is

determined by the number of sarcomeres in series, and tendon length (Zajac 1989); however, at the level of the sarcomere there is a single unique active force-length relationship for human skeletal muscle, which consists of an ascending limb, plateau region, and descending limb (Walker and Schrodt, 1974; Gollapudi and Lin, 2009).

The plateau region of this relationship, at lengths between ~2.6 and 2.8 μm in human skeletal muscle, is optimal for active force production; it is this length at which a maximum number of myosin head groups can form cross-bridges with actin. Sarcomere length operating range is a function of each of muscle length, fibre length, and the geometric orientation of the muscle (moment arm) with respect to the joint(s) that it crosses. This operating range can be considered the range each sarcomere will shorten or lengthen through normal joint motion. By having short sarcomere lengths in a neutral posture and a narrow operating range in spine flexion, multifidus force production is inherently limited in a neutral spine position and can only produce maximal isometric force when the spine is in full flexion

(Ward et al., 2009). On the contrary, psoas major has long sarcomeres in a neutral spine position and approaches optimal lengths during full hip flexion (Regev et al.,

2011). The effect of lumbar spine position on sarcomere length operating ranges of psoas major and the remaining lumbar muscles are unknown.

The purposes of this study were to: (1) measure neutral cadaveric sarcomere lengths of the muscles that surround the lumbar spine; and (2) model the sarcomere length operating range through full lumbar spine flexion, extension, lateral bending, and axial rotation. As multifidus and psoas major lie posterior and anterior to the spine, respectively, the lumbar extensor (posterior) muscles were hypothesized to

operate on the ascending limb of the force-length relationship, similar to multifidus, while muscles located more anteriorly would have operating ranges more similar to psoas major on the descending limb.

3.3 Materials and Methods

3.3.1 Cadaveric Donors

Muscle sarcomere lengths were measured from one side of the body of seven cadaveric donors (6 male, 1 female, aged 77 ± 8.5 years). All donors were embalmed in an approximately neutral anatomical posture and had no visible musculoskeletal or spine-related injuries. The embalmed nature of the muscles is necessary for the type of architectural measurements made in this study (Lieber and

Fridén, 2000) to prevent muscle fibres from shortening under their own passive tension when they are excised from the skeleton. The Research Ethics Board at the university approved the use of the cadavers for this study.

Surrounding tissues were carefully dissected to expose the muscle fascicles, tendons, and bony insertions of the lumbar spine muscles. Samples approximately

10-12 fascicles thick and 0.5 cm long were excised from muscles of interest

(Table 3.1) and stored in phosphate-buffered saline solution. After sampling, the remaining musculature was removed to expose deeper tissues. A posterior approach was used to sample iliocostalis, longissimus, multifidus, and intertransversarii, while the psoas muscles and quadratus lumborum were accessed anteriorly. Psoas minor was absent in four of the seven donors, resulting in three psoas minor muscles examined. Intertransversarii muscles from one donor were

not fully embalmed and were removed from analysis, resulting in intertransversarii muscles of six donors being examined.

Table 3.1: Location and attachment sites of all muscles fascicles that were sampled. Samples were taken in the approximate mid-belly of the muscle between attachment sites. Muscle Location Inferior attachment Superior attachment Iliocostalis† Thoracic Iliac crest via ESA Angle of rib 8 Lumbar Iliac crest L1 transverse process Longissimus† Thoracic Iliac crest via ESA Angle of rib 4 Lumbar Iliac crest L1 transverse process Multifidus L1 Posterior superior iliac spine L1 spinous process L4 Sacrum L4 spinous process Intertransversarii Medialis L2-L5 mammillary process L1-L4 accessory process Lateralis L2-L5 transverse process L1-L4 transverse process Psoas major L1 Pectineal line of pelvis L1 vertebral body L4 Lesser trochanter of femur L4 vertebral body Psoas minor L1 Pectineal line of pelvis L1 vertebral body Quadratus lumborum Iliac crest Rib 12 ESA: erector spinae aponeurosis †In the lumbar region, the intramuscular septum was used to divide iliocostalis and longissimus (Bogduk 1980). Thoracic and lumbar samples all arose from the iliac crest directly or indirectly via the ESA. Thoracic level iliocostalis samples were therefore taken from iliocostalis lumborum rather than iliocostalis dorsi (Gray 1985).

3.3.2 Sarcomere length measurement

Laser diffraction, using the methods developed by Lieber et al. (1990) and described in Chapter 2: Experiment I, was used to measure sarcomere lengths of three to six different fascicles for each muscle sample. Each sarcomere length measurement yields an average value from thousands of individual sarcomeres that fall within the diameter of the laser beam on the muscle fascicle.

3.3.3 Modelled operating ranges

Operating ranges of lumbar spine muscles were modelled using anatomically detailed geometric coordinate data, representing a 50th percentile male, from the model of Cholewicki & McGill (1996). Muscles were modelled as lines of action connecting skeletal attachment sites (e.g. specific bony locations, representing muscle origin and insertions, on the pelvis and spine; the erector spinae line of action also included a nodal point to account for curvature along the spine) using a custom MATLAB 7.0 program (MathWorks, Natick MA) (Figure 3.1). The superior muscle attachments were rotated sequentially about each intervertebral joint from

L5/sacrum to L1/L2, accounting for differences in range of motion of each level, throughout the population average full range of lumbar flexion, extension, and both lateral bend and axial rotation ipsilaterally and contralaterally (Table 3.2; White and

Panjabi 1990; McGill 2007). Therefore, rotated muscle origins are dependent on the number of intervertebral joints crossed, the distance from the joint centre, and the magnitude of rotation at each joint. The total lengths of each vector, representing individual muscle fascicles, were calculated in both neutral and rotated positions.

The percent change ([rotated length – neutral length]/neutral length*100) was used to predict the sarcomere length at the end range of each motion. Uniform stretching of sarcomeres along each fascicle was assumed. While it is known that sarcomere strains are not uniform during whole muscle stretching (Moo et al., 2016), this assumption is accurate for defining the average SL along a muscle fibre; the change in muscle fibre length must be equal to the sum of the changes of sarcomere length of all sarcomeres along a fibre.

∆�! = ∆��! = � ∗ ∆�� !!!:! where LM is muscle length, SL is the sarcomere length and n is the number of sarcomeres. For the intertransversarii, despite measuring sarcomere lengths at multiple intervertebral levels, only the L4-L5 level was modelled, as coordinate data for the intertransverse ligament were only available at this site. As this ligament and the intertransversarii muscle occupy essentially the same anatomical space, their common coordinates can be used for this purpose. Longissimus and iliocostalis were modelled as one muscle (erector spinae), as coordinate data were only available for this combined muscle group. Finally, the thoracic regions of longissimus and iliocostalis muscles were not modelled, as the coordinate data only allowed for the assessment of lumbar spine range of motions. Spinalis was not modelled, as it does not insert onto the iliac crest of the pelvis and is mostly aponeurotic in the lumbar spine (Bogduk 1980).

Table 3.2: Modeled intervertebral rotations, taken as population averages (White and Panjabi, 1990; McGill 2007).

Flexion Extension Lateral Bend Axial Rotation Total Rotation 52.0° 16.1° 29.0° 13.1° Level L1 8.3° (16%) 5.0° (31%) 6.1° (21%) 2.0° (15%) L2 9.9° (19%) 3.0° (19%) 6.1° (21%) 2.0° (15%) L3 12.0° (23%) 1.0° (6%) 7.8° (27%) 2.1° (16%) L4 13.0° (25%) 2.1° (13%) 6.1° (21%) 2.1° (16%) L5 8.8° (17%) 5.0° (31%) 2.9° (10%) 4.9° (38%)

3.3.4 Statistical Analysis

A mixed model repeated measures ANOVA (fixed effect: muscle; mixed effects: donor and donor*muscle) was used to determine if muscles were different in their measured cadaveric neutral spine sarcomere lengths. Contrast statements were run to compare different levels of the same muscle. A two-way mixed model repeated measures ANOVA was used to determine effects of muscle (medialis and lateralis) and intervertebral level (L1/L2 to L4-L5) on the sarcomere length of intertransversarii. Analysis of residuals determined that residuals were normally distributed and random. A Tukey (HSD) test was used to account for multiple comparisons. SAS 9.3 (SAS Institute Inc., Cary NC) was used for all statistical tests.

3.4 Results

3.4.1 Neutral Spine Cadaveric Sarcomere Lengths

Sarcomere lengths were significantly different (p < 0.001) between muscles.

The post-hoc comparison of all muscles determined that there were no differences between any of the extensor muscles (longissimus, iliocostalis, and multifidus) and that all of these muscles had shorter than optimal sarcomere lengths for active force generation (Table 3.3). The average lumbar multifidus sarcomere length

(2.36 ± 0.05 μm) was located on the ascending limb of the sarcomere force-length curve. The longissimus and iliocostalis sarcomere lengths were also located on the ascending limb (Table 3.3). There were no differences in sarcomere lengths between the lumbar multifidus fibres attaching to the L1 and L4 spinous processes

(p = 0.3929). The pooled iliocostalis and longissimus sarcomere lengths were statistically different (p = 0.0334) between the thoracic (2.50 ± 0.06 μm) and lumbar regions (2.35 ± 0.05 μm).

Table 3.3: Sarcomere lengths (μm) of all muscles measured from an approximate neutral spine position in cadaveric donors. Muscle Level Mean SE Multifidus 2.36 0.05 A

L1 2.32 0.06 H0: L1 = L4 L4 2.39 0.08 p = 0.3929

Iliocostalis 2.41 0.06 A

Lumborum 2.34 0.07 H0: Lumb = Thor Thoracis 2.48 0.11 p = 0.1325

Longissimus 2.43 0.05 A

Lumborum 2.36 0.09 H0: Lumb = Thor Thoracis 2.51 0.06 p = 0.1275

Intertransversarii 2.53 0.05 AB Lateralis L1-L2 2.55 0.08 H0: Lat = Med L2-L3 2.52 0.05 p = 0.6373 L3-L4 2.43 0.07 L4-L5 2.53 0.04 Medialis L1-L2 2.48 0.08 Effect of Level † L2-L3 2.64 0.12 p = 0.5801 L3-L4 2.53 0.13 L4-L5 2.60 0.06

Quadratus lumborum 2.66 0.09 B

Psoas minor 3.17 0.12 C

Psoas major 3.41 0.07 D

L1 3.36 0.09 H0: L1 = L4 L4 3.46 0.10 p = 0.2465 Different letters (A-D) indicate post-hoc statistically different sarcomere lengths (p < 0.05) between muscles. Bolded values are the averages sarcomere lengths across all sites for each muscle. HO: designate the null hypothesis for each contrast statement. † F-test for the statistical significant variation between levels in intertransversarii

The quadratus lumborum, located intermediately between the extensor and anterior muscles, was the only muscle measured in the optimal region for active force production (2.66 ± 0.09 μm). Quadratus lumborum was statistically different from all muscles (p < 0.05), except the similarly located intertransversarii muscles

(p = 0.1390), which were near optimal length (2.53 ± 0.05 μm). There was no statistical difference between intertransversarii medialis and lateralis (p = 0.6373), nor was there an effect of vertebral level (p = 0.5801).

The anteriorly located psoas major and minor muscles were found on the descending limb of the force-length relationship, with psoas major having longer sarcomeres than psoas minor (3.41 ± 0.07 and 3.17 ± 0.12 μm, respectively; p = 0.0306). Both muscles had longer sarcomeres than all other muscles

(p < 0.0001).

3.4.2 Modelled Sarcomere Lengths

3.4.2.1. Flexion and Extension

In modelled lumbar spine flexion of 52°, the largest relative length change of

50.0% was found in the multifidus fibres originating from L1 (Table 3.4). The sarcomeres lengthened past optimal and down the descending limb of the force- length relationship to 3.48 μm at full lumbar spine flexion (Figure 3.2A). The shorter multifidus fibres, originating at the L4 spinous process, had a narrower operating range with a final length of 3.11 μm.

Table 3.4: Modeled sarcomere lengths (µm) and percent change (below in brackets) from the neutral posture for each muscle of interest at end ranges of lumbar spine motion. Muscle Neutral* Flexion Extension Lateral Bend Axial Rotation Ipsilateral Contralateral Ipsilateral Contralateral Multifidus L1 2.32 3.48 1.97 2.16 2.46 2.41 2.26 (50.0) (-15.1) (-6.7) (6.0) (4.0) (-2.5) L4 2.39 3.11 2.29 2.38 2.40 2.41 2.38 (30.0) (-4.1) (-0.4) (0.4) (0.7) (-0.6) Iliocostalis lumborum† 2.34 3.23 2.09 2.03 2.65 2.40 2.32 (38.1) (-10.5) (-13.3) (13.3) (2.6) (-0.8) Longissimus lumborum† 2.36 3.26 2.11 2.05 2.67 2.42 2.34 (38.1) (-10.5) (-13.3) (13.3) (2.6) (-0.8) Intertransversarii‡ 2.53 3.11 2.45 2.22 2.85 2.53 2.54 (23.3) (-3.3) (-12.3) (12.7) (0.0) (0.3) Quadratus lumborum 2.66 3.44 2.46 2.00 3.29 2.76 2.59 (29.2) (-7.2) (-24.9) (23.7) (3.7) (-2.5) Psoas minor L1 3.17 3.12 3.12 2.98 3.33 3.19 3.18 (-1.7) (-1.7) (-6.2) (4.8) (0.3) (0.1) Psoas major L1 3.36 3.30 3.30 3.15 3.52 3.37 3.36 (-1.7) (-1.7) (-6.2) (4.8) (0.3) (0.1) L4 3.46 3.46 3.46 3.44 3.49 3.46 3.47 (0.0) (0.0) (-0.5) (0.9) (-0.1) (0.3) Values in brackets are percent change in sarcomere length compared to the neutral spine posture calculated as [(Rotated length – Neutral length)/Neutral length]*100%. * Measured rather than modelled lengths † Iliocostalis and longissimus were only modelled in the lumbar region. ‡ Intertransversarii was only modelled at the L4-L5 level

The amount of sarcomere lengthening decreased as attachment sites

travelled laterally and anteriorly from the spinous processes around to the anterior

vertebral body. For iliocostalis and longissimus, only a single erector spinae

attachment site on both the iliac crest and transverse process was modeled,

resulting in the same predicted length change of 38.1%, which is slightly greater

than the 29.2% length increase of quadratus lumborum. The psoas major and minor

fibres attaching to L1 were largely unaffected by lumbar spine flexion resulting in

1.7% shorter sarcomeres. The shorter psoas major fibres attaching to L4 were not

affected by lumbar spine flexion.

Intertransversarii were modelled at L4-L5 because the only available coordinate data were for the L4-L5 intertransverse ligament. This muscle experienced a 23.3% increase in sarcomere length during flexion.

At full lumbar spine flexion, all measured sarcomeres were located on the descending limb of the force-length relationship, and values ranged from 3.11 μm for L4-L5 intertransversarii and L4 multifidus, to 3.48 μm for L1 multifidus

(Figure 3.2A).

During modelled lumbar spine extension of 16°, the same pattern was observed with the largest relative length change occurring in the L1 multifidus, followed sequentially by muscles attaching further laterally and anteriorly. For the muscles originating on the pelvis and attaching to L1, multifidus shortened by

15.1%, the erector spinae by 10.5%, and quadratus lumborum by 7.2%. The shorter

L4 multifidus crosses fewer intervertebral levels and was consequently only 4.1% shorter during extension. The L1 psoas major and minor shortened by 1.7%, while the L4 psoas major was unaffected. The L4-L5 intertransversarii shortened by

3.3%.

3.4.2.2 Lateral bending

During lumbar spine lateral bending (29°), the muscle with the largest operating range was quadratus lumborum. Quadratus lumborum became 24.9% shorter ipsilaterally and 23.7% longer contralaterally. Muscles that attached more

medially on the vertebrae (psoas major, psoas minor, and multifidus) had shorter operating ranges (Figure 3.2B).

3.4.4.3 Axial Rotation

In the lumbar spine, axial rotation (13°) had very little effect on the sarcomere lengths of lumbar muscles. Quadratus lumborum and the multifidus attaching to L1 were the most affected, becoming 3.7% and 4.0% longer on the ipsilateral side, respectively. Each became 2.5% shorter contralaterally

(Figure 3.2C).

3.5 Discussion:

The sarcomere length operating ranges illustrate how the muscles of the spine are designed to work together as a highly effective system, which balance each other by inherently modifying their force generating capacity through the lumbar spine range of motion. In the neutral spine posture, posterior extensor muscles had short sarcomeres on the ascending limb of the force-length relationship, anterior muscles had long sarcomeres on the descending limb, and intermediate muscles had optimal, or close to optimal, lengths (Figure 3.3A). During modelled spine flexion, the anterior muscle sarcomeres remained relatively unchanged, while the extensor and intermediate muscle sarcomeres lengthened past optimal. Thus, all muscles achieved similar non-optimal sarcomere lengths at full lumbar spine flexion, where active force production would be limited to sub-maximal levels (Figure 3.3B). While

active force production is inherently limited, all muscles had long sarcomere lengths and would be predicted to generate large passive forces in this position.

The hypothesis that other posterior extensor muscles would be similar to multifidus was accepted. There were no differences found between the intrinsic sarcomere length of multifidus, longissimus, and iliocostalis in either neutral or modelled postures. Because the extensor muscles have short sarcomeres in a neutral posture, these muscles become inherently stronger as the spine flexes forward, a position known to increase the vulnerability of the spine to injury

(Andersson et al., 1977; Adams and Hutton, 1982; Snook et al., 1998; Gunning et al.,

2001). This creates a proportional force response for spine movement, where greater displacement from a neutral posture creates a greater potential for restoring the spine to a neutral position. As the extensor muscles have the same intrinsic sarcomere length design, these muscles do not serve completely isolated functions, but rather work in synergy. Synergistic muscle systems have been shown to be more effective than a single super muscle in other areas of the body (Lieber and

Fridén, 2000). For example, extensor carpi radialis brevis (ECRB) and longus

(ECRL) create a synergistic system, where ECRB has a greater PCSA and ECRL has longer fibres. If ECRB and ECRL were replaced with a single muscle, capable of producing the combined force of both muscles while maintaining the larger excursion of ECRL, the new muscle would need to be 30% larger. Similar to the wrist extensors, multifidus and the erector spinae muscles also have different architectures, with multifidus having a large PCSA and erector spinae having greater moment arms and fascicle lengths (McGill, 1991; Ward et al., 2009; Delp et al., 2001;

Rosatelli et al., 2008). This enables the multifidus to generate relatively large forces and the erector spinae to generate more moderate forces, but still substantial

moments, over a greater range of vertebral levels. These divergent architectures further unite through a similar design at the level of the sarcomere and thus work in synergy to create a highly effective system to actively produce force in a semi-flexed spine posture.

Multifidus sarcomere lengths measured in the neutral spine position were similar to the cadaveric lengths previously published by Ward et al. (2009).

However, the cadaveric measurements are longer than the intraoperative measurements found in chronic low back pain patients (Ward et al., 2009). This may not be surprising because cadaveric measurements from fixed biopsies are not always the same as those measured in situ (Lieber et al., 1994). This variability could result from differences in cadaveric and intraoperative spine position or tissue stretching due to hydration of embalmed cadavers. Regardless of the cause of the differences, two possibilities arise: (a) our data could overestimate the neutral and modelled sarcomere lengths; or (b) the spine in Ward et al. (2009) may not have been positioned in full flexion intraoperatively. Overestimation of the neutral and modelled sarcomere lengths would mean that in reality the extensor muscles are even more limited in their active force generating capability in a neutral posture, and would work at optimal lengths when the spine is fully flexed. This may seem to be a practical scenario where the muscles are able to produce the greatest amount of restorative active extensor force in a position in which the lumbar spine is highly vulnerable to injury and requires the greatest moment for support. The second possibility that the spine was not positioned in full flexion intraoperatively in Ward et al. (2009) suggests that our predictions could be most accurate and thus the

extensor muscles reach a point mid-flexion where their active force production begins to decrease. While less intuitive, this may not be a flaw in the muscle design.

As the spine flexes forward, passive tissues such as muscles, ligaments, and the intervertebral discs become strained. Experiment I demonstrated that passive muscles generate considerable stresses when strained to long lengths.

Consequently, spine extensor muscles are not required to generate large active forces in fully flexed postures. This has been demonstrated by the well known flexion-relaxation phenomenon, where posterior extensor muscles become inactive as the spine approaches full flexion (Floyd and Silver, 1955; McGill and Kippers,

1994; Toussaint et al., 1995). It appears that spine muscles are functionally well designed to produce large extensor moments throughout the range of spine motion; passive muscle components generate large forces in fully flexed postures while active components generate the greatest forces in mid flexion.

The lumbar spine muscles have smaller sarcomere operating ranges in both lateral bend and axial rotation than flexion/extension. This is not surprising as lateral bend and axial rotation have less range of motion than flexion/extension.

These operating ranges are centered on the neutral sarcomere length of each muscle due to their side-to-side symmetry. Unlike flexion and extension where posterior and anterior muscles balance each other, in axial rotation and lateral bend muscles are inherently balanced by their contralateral counterparts.

An important consideration for the predicted sarcomere length operating ranges is that our model assumes that length changes occur uniformly throughout each muscle-tendon unit. During passive movements muscle is more compliant

than tendon (Stolov and Weilepp 1966). If tendon comprised a significant portion of the muscle-tendon unit length, then the compliant muscle would stretch to a relatively greater extent and the sarcomere operating ranges modelled here would be underestimated. However, this is likely not a large source of error in this study.

Multifidus, quadratus lumborum, and intertransversarii have negligible tendons compared to the length of their muscle fibres. The fibres of longissimus and iliocostalis attaching to the lumbar transverse processes (modeled here) also have negligible tendons, as opposed to the fibres attaching to the ribs (which we did not model). The psoas major and minor muscles have longer tendons, which could lead to differences in relative lengthening between tendon and muscle. However, lumbar spine motion generates fairly small changes in overall length of psoas muscle- tendon units. Therefore relative differences in muscle and tendon lengthening for psoas major and minor would have little effect on changes in sarcomere length.

It is important to note that we measured and modeled muscle sarcomere lengths, which define maximum isometric sarcomere force generating capability.

Muscle physiological cross-sectional area and moment arms are very important determinants for absolute muscle force and moment generating capability, respectively. Further, neural drive modifies muscle force production. All of these factors will influence how muscles interact synergistically to move, load, and stabilize the spine. Specifically, here we have demonstrated that from a sarcomere organizational perspective, it appears that the lumbar spine muscles have been designed to act together to balance the capability for active force generation as the lumbar spine moves through full flexion.

The arrangement of sarcomeres within muscles appears to provide an inherent mechanism for balancing force production throughout spine motion.

Changes in sarcomere length, in relation to changes in spine posture, modify the maximum capability of muscles to produce active force. Muscles within the same region have similar sarcomere operating ranges, implying that they are designed to work together to generate optimal levels of force across the same ranges of spine motion. Specifically, the posterior extensor muscles are all designed to produce maximum active force in a mid-flexed posture, and all muscles are designed to have long sarcomere lengths in a fully flexed posture. These long sarcomere lengths hinder active force production, however are also associated with large passive forces.

3.6 Bridge Summary

Combining the results of Experiments I and II allows the musculoskeletal model to predict passive and maximally activated muscle forces based on spine posture. Currently, measuring the position and orientation of each lumbar vertebrae is very difficult during dynamic human movements, due to the number of degrees of freedom of movement and the large amount of soft tissues surrounding the vertebrae. For this reason total lumbar (ie sacrum to T12) flexion/extension, lateral bend, and axial twist angles are recorded using kinematics and the percentage of rotation at each joint is predicted using population averages. We are currently working on new techniques to record intersegmental spine angles in order to further improve spine modelling; however, this work is ongoing.

Before the musculoskeletal model can be used to predict muscle activation patterns during dynamic trunk flexion, kinematic and kinetic data sets need to be recorded as input for the model. Muscle activation patterns also need to be recorded during these movements in order to validate the predicted activations. The generation of these data sets was one of the goals of Experiment III.

Chapter 4: Experiment III

Decreasing the required lumbar extensor moment induces earlier onset of flexion relaxation

4.1 Chapter summary

Flexion relaxation is characterized by the lumbar erector spinae becoming myoelectrically silent near full trunk flexion. This study was designed to: 1) determine if decreasing the lumbar moment during flexion would induce flexion relaxation to occur earlier; 2) characterize thoracic and abdominal muscle activity during flexion relaxation. Ten male participants performed four trunk flexion/extension movement conditions; lumbar moment was altered by attaching

0, 2.27, 4.54, or 6.80 kg counterweights to the torso. Electromyography (EMG) was recorded from eight trunk muscles. Lumbar moment, lumbar flexion and trunk inclination angles were calculated at the critical point of lumbar erector spinae inactivation. Results demonstrated that counterweights decreased the lumbar moment and lumbar flexion angle at the critical point of inactivation (p<0.0001 and p=0.0029, respectively); the hypothesis that flexion relaxation occurs earlier when lumbar moment is reduced was accepted. The counterweights did not alter trunk inclination at the critical point of inactivation (p = 0.1987); this is believed to result from an altered hip to spine flexion ratio when counterweights were attached.

Lumbar multifidus demonstrated flexion relaxation, similar to lumbar erector spinae, while thoracic muscles remained active throughout flexion. Abdominal muscles activated at the same instant as lumbar erector spinae inactivated, except in the 6.80 kg condition where abdominal muscles activated before lumbar erector spinae inactivated resulting in a period of increased co-contraction. 63

4.2 Introduction

Flexion relaxation is characterized by a reduction in spine extensor muscle activation during trunk flexion that generally occurs in individuals free from back pain (Allen, 1948; Floyd and Silver, 1951; 1955). While flexion relaxation reliably occurs in healthy individuals, there is evidence that flexion relaxation does not occur in patients suffering from chronic low back pain; this population maintains muscle activity throughout the full trunk flexion range of motion (Golding, 1952; Ahern et al., 1988; Mannion et al., 2001). Clinically, the absence of flexion relaxation is currently used as an objective measure of low back pain (Neblett et al., 2003), and chronic pain interventions use flexion relaxation as a goal for biofeedback retraining

(Neblett et al., 2010; Moore et al., 2015). However, a better understanding of the factors that affect flexion relaxation is warranted in order to apply this technique effectively in a clinical population (Geisser et al., 2005).

The proposed mechanism for flexion relaxation is that passive tissues that are slack in a neutral spine position become stretched during spine flexion and begin to resist the externally applied moments about lumbar joints (Pauly, 1966;

Wolf, 1979; McGill and Kippers, 1994). This increase in passive tension thereby reduces the required level of active force production in extensor spine muscles leading to full cessation of the electromyographic (EMG) signal. Flexion relaxation is most commonly observed in the lumbar erector spinae at the L3/L4 spine level

(Floyd and Silver, 1951; Pauly, 1966; Shirado et al., 1995; Solomonow, 2003; Olson et al., 2006; Shin and Mirka, 2007; Jin et al., 2012; Hashemirad et al., 2009; Schinkel-

Ivy et al., 2013), but has also been observed at the L2/L3 (Allen, 1948; Golding,

1952; Morin and Portnoy, 1956; Morris et al., 1962; Shin et al., 2004; Olson et al.,

2004) and L4-L5 spine levels (Wolf et al., 1979; Golding, 1952; Dickey et al., 2003;

Olson et al. 2004). The passive spine structures are commonly thought to generate the lumbar extensor moment, yet the thoracic erector spinae and latissimus dorsi may also actively contribute to the lumbar moment, thereby reducing the active requirements of the lumbar erector spinae. Floyd and Silver (1955) mention that thoracic erector spinae also exhibits flexion relaxation, although later studies found that thoracic erector spinae maintained activity during full trunk flexion (Morris et al., 1962; Pauly, 1966). Indwelling EMG has been used to determine that lumbar multifidus also demonstrates flexion relaxation (Morris et al., 1962; Pauly, 1966;

Donish and Basmjian, 1972). While surface EMG recordings of lumbar erector spinae commonly show flexion relaxation, indwelling EMG of iliocostalis lumborum has shown that activity is maintained during full trunk flexion (Andersson et al.,

1996); quadratus lumborum also maintained activity in the same study. The abdominal muscles have largely been ignored in studies of flexion relaxation, although they have been suspected of being responsible for generating the moment to achieve the last few degrees of trunk flexion after the extensor muscles have become inactive (Solomonow, 2003; Olson et al., 2004).

The instant the lumbar erector spinae muscles become inactive is termed the critical point of flexion relaxation and is the most common outcome measure in the analysis of flexion relaxation. Often the lumbar flexion angle—either in absolute degrees (Gupta, 2001; Shin and Mirka, 2007; Hu et al., 2013) or normalized to

maximum lumbar flexion (Kippers and Parker, 1984; Sarti et al., 2001; Schinkel-Ivy et al., 2014)—measured at the critical point is compared between conditions. Both the L4-L5 moment and trunk inclination angles—thoracic spine segment with respect to global reference system— at the critical point have also been emphasized as important measures of flexion relaxation (Olson et al., 2004; Howarth and

Mastragostino, 2013; Zwambag and Brown, 2015). Repeated trunk flexion and the addition of loads to the trunk or hands have been shown to delay flexion relaxation in healthy participants, resulting in greater flexion angles (Dickey et al., 2003;

Solomonow, 2003) and L4-L5 extensor moments (Howarth and Mastragostino,

2013) at the critical point. It is believed that flexion relaxation is delayed in these studies by decreasing the capability of passive tissues to generate extensor moments and by increasing the required extensor moment, respectively. These findings therefore corroborate the proposed mechanism that flexion relaxation occurs when the passive tissues are able to support the lumbar moment; as expected, adding loads to the hands or trunk increased the L4-L5 moment and delayed flexion relaxation. This mechanism would also suggest that flexion relaxation should occur earlier if the L4-L5 moment was reduced; however, this remains to be determined.

The primary purpose of this study was thus to investigate how reducing the

L4-L5 moment affected the onset of flexion relaxation. It was hypothesized that counterweights acting through a pulley attached to the thoracic spine would reduce the L4-L5 moment. Consequently, less passive tissue strain would be required to equilibrate the external moment and the critical point for lumbar erector spinae

inactivation would occur earlier in the flexion movement with less lumbar flexion and trunk inclination. A secondary purpose of this study was to characterize the activity of the surrounding musculature during trunk flexion. Abdominal muscles were hypothesized to become activated after the lumbar erector spinae become inactive and that their activity would be increased with larger loads attached to the pulley. Lumbar and thoracic erector spinae, latissimus dorsi, and multifidus were hypothesized to demonstrate a reduction in activity throughout trunk flexion as larger loads were attached to the pulley.

4.3 Methods

4.3.1 Participant characteristics

Ten healthy male participants (mean ± SD; age: 25 ± 2.5 years; height: 181 ±

5.8 cm; mass: 82 ± 11.2 kg) were recruited from the university. Participants had no previous history of low back pain. The research ethics board at the university approved this study.

4.3.2 Experimental set-up

Standard bipolar Ag/AgCl surface electrodes (Blue Sensor, Medicotest Inc.,

Ølstykke, Denmark) were used to record bilateral muscle activations from lumbar and thoracic erector spinae, latissimus dorsi, rectus abdominus, external oblique, and internal oblique. Electrodes were placed along muscle fibre directions at L3, T9, and T12 spine levels for lumbar and thoracic erector spinae and latissimus dorsi, respectively. Electrodes were placed along muscle fibres ~2 cm lateral to the naval

for rectus abdominus, ~14 cm lateral to the midline for external oblique, and ~2 cm medial and inferior to the anterior superior iliac spines for internal oblique (Brown and McGill, 2010). These electrode sites have been shown to adequately reflect abdominal wall activation and reduce cross-talk (McGill, 1996). Ground electrodes were placed over the anterior superior iliac spines; all electrode sites were shaved, if necessary, and cleaned with isopropyl alcohol. Muscle activations of multifidus at both the L1 and L4 spine levels were recorded using fine-wire EMG. Multifidus EMG was limited to the right side to avoid obstructing the line of sight of kinematic markers (described later). Bipolar 44 μm gauge fine wire nickel alloy electrodes with 2 mm exposed tips bent into hooks (50 mm x 25 gauge, Chalgren Enterprises

Inc., Gilroy, CA, USA) were inserted into the multifidus muscle with a 27 gauge hypodermic needle, ~1.5 cm lateral to the L1 and L4 spinous processes in a slight craniomedial orientation. Prior to needle insertion, multifidus muscle thicknesses

(mean ± SD; 3.2 ± 0.50 and 3.9 ± 0.54 cm at L1 and L4, respectively) were measured as the distance from erector spinae aponeurosis to lumbar laminae using ultrasound

(M-Turbo, Sonosite Inc., Bothell, WA, USA). This was to ensure that needles were inserted into the middle of the muscle bellies. Electromyographic data were differentially amplified (AMT-8, Bortec Biomedical, Calgary, Canada; bandwidth 10-

1000 Hz; common-mode rejection ratio = 115 dB at 60 Hz; input impedance = 10

GΩ) and recorded at 2048 Hz.

Participants performed three maximal voluntary isometric contractions

(MVICs) targeting each muscle group. For lumbar and thoracic erector spinae, and multifidus, participants adopted the prone Beiring Sørensen position and manual

resistance was applied against the generated trunk extensor moment (Vera-Garcia et al., 2006). For latissimus dorsi, participants stood and performed a rowing

(humeral extension) action while an experimenter provided manual resistance

(Beaudette et al., 2014). For abdominal muscles participants sat with knees bent and trunk slightly reclined on the edge of a bench and manual resistance was applied as participants generated moments about each of trunk flexion, left and right axial twist, and left and right lateral bend.

Rigid bodies consisting of two kinematic markers (Optotrack, NDI Inc.,

Waterloo, Canada) were placed over the spinous processes of T12 and S1. Single markers were attached to the head of the 5th metatarsal, lateral malleolus, lateral knee, and greater trochanter on the participants’ left side. Kinematic data were recorded at 32 Hz. Participants stood on a force plate (AMTI Inc., Watertown, USA); ground reaction forces and moments were amplified and collected at 2048 Hz.

A harness was strapped to the participants’ torso allowing a cable and fixed pulley to be connected to the trunk at approximately the T5 spine level. The pulley was mounted overhead so that masses attached to the cable generated an extensor moment about the lumbar spine (Figure 4.1). Therefore, any mass attached to the pulley acted to reduce the flexion moment of the torso.

4.3.3 Protocol

A repeated measures randomized block design was used for this study. For every trial, participants were instructed to stand in a neutral position, flex forward to full trunk flexion over a period of three seconds, hold this position for three

seconds, and then return to upright stance over a period of three seconds.

Participants were asked to reach maximum trunk flexion in all trials; no explicit instructions were given on how to achieve this. Zero, 2.27, 4.54, or 6.80 kg masses were attached to the pulley in a random order to generate four different flexion/extension conditions; each participant performed three blocks containing all four conditions. One minute and five minutes of rest were provided between each trial and block, respectively, to reduce the likelihood of fatigue and viscoelastic changes in tissue properties.

4.3.4 Data Processing

Surface EMG data were first high-pass filtered (4th order dual-pass

Butterworth) with an effective cut-off of 100 Hz to reduce heart rate contamination

(Staudenmann et al., 2007; Moreside et al., 2008) and improve representation of predicted force profiles (Potvin & Brown 2004). EMG data were then de-biased, linear enveloped (rectified and low-pass filtered with a 2nd order Butterworth at 2.5

Hz), and expressed as a percentage of maximal activity from the MVIC trials. Critical points of muscles turning ‘on’ (lumbar erector spinae) or ‘off’ (rectus abdominus and external oblique) were determined using a threshold of mean + 2*SD of an identified quiet period; signals must cross threshold for >20ms (Howarth and

Mastragostino, 2013). The identified quiet periods for lumbar erector spinae and abdominal muscles occurred during full trunk flexion and quiet standing, respectively. All critical points were verified manually. Critical points were not calculated for thoracic erector spinae, latissimus dorsi, multifidus, and internal

oblique as these muscles did not have identified quiet periods and therefore did not turn ‘on’ or ‘off’. The L4-L5 moment, lumbar flexion angle, and trunk inclination angle corresponding to each critical point were calculated. To compare muscle activities between conditions, each trial was time normalized to maximum lumbar flexion angle and averaged across participant

Kinematic and kinetic data were low-pass filtered (4th order dual-pass

Butterworth) with an effective cut-off of 3 Hz and kinetic data were down sampled to 32 Hz. Kinematic and kinetic data were then used as input to a custom bottom-up, dynamic, two-dimensional, rigid linked segment model to estimate the moment at the L4-L5 intervertebral disc; mass moments of inertia, locations of centre of mass, and masses of each segment were taken from Winter (2009) and Zatsiorsky (1998).

Lumbar flexion angle was calculated as the angle between T12 and S1 rigid bodies in the sagittal plane, and trunk inclination was calculated as the angle between the T12 rigid body and the global horizontal axis (Figure 4.1). Lumbar flexion and trunk inclination angles in a neutral standing posture were set to zero.

4.3.5 Statistical Analysis

Temporal differences between critical points of lumbar erector spinae and abdominal muscles were statistically tested using analyses of variance (ANOVA).

Differences between the magnitudes of L4-L5 moments, lumbar flexion angle, and trunk inclination angle at critical points were also tested between conditions (0,

2.27, 4.54, and 6.80 kg) using repeated measures ANOVA. Tukey’s Honest

Significant Difference test for multiple comparisons was used; significance was

determined using an α of 0.05. Activation patterns of all muscles were additionally time-normalized and compared qualitatively.

4.4 Results

4.4.1 Lumbar Erector Spinae

Data were pooled between left and right muscles because there were no differences in the critical point for lumbar erector spinae inactivation between sides

(F = 2.10; p = 0.1497). Mass added to the pulley caused the lumbar erector spinae to inactivate at a lower L4-L5 moment (F = 83.06; p = <0.0001). In the 0 kg condition, the moment about the L4-L5 disc at the critical point of lumbar erector spinae inactivation was 93.7 ± 2.5 Nm. Adding 2.27, 4.54, and 6.80 kg to the pulley reduced the L4-L5 moment at the critical point to 78.9 ± 2.4, 66.5 ± 2.4, and 61.4 ± 2.2 Nm, respectively (Figure 4.2).

Figure 4.2: Mean (+ SEM) L4-L5 moment, lumbar flexion and trunk inclination angles at the critical point of lumbar erector spinae inactivation. The L4-L5 moment and lumbar flexion angles decrease as greater masses are attached to the pulley. There was no effect of condition on trunk inclination. Different letters indicate post- hoc statistically different means between conditions.

Lumbar erector spinae inactivation also occurred with less lumbar flexion when the L4-L5 moment was reduced by the counterweights (F = 10.29; p =

0.0029). In the 0 kg condition, flexion relaxation occurred at 51.8 ± 1.2° of lumbar spine flexion. Flexion relaxation occurred at 50.8 ± 1.2, 49.7 ± 1.1, and 48.6 ± 1.2° of lumbar spine flexion when 2.26, 4.54, and 6.80 kg were attached to the pulley, respectively (Figure 4.2). Trunk inclination at the critical point of lumbar erector spinae inactivation was not affected by the counterweights (F = 1.91; p = 0.1987); the critical point occurred at 100.4 ± 0.7° of trunk inclination (Figure 4.2).

The decreased lumbar moment and lumbar flexion angle at the critical point indicates that the lumbar erector spinae became inactive earlier in the trunk flexion movement. The time-normalized trace (Figure 4.3) also provides evidence that the lumbar erector spinae muscle was less active throughout the trunk flexion movement as greater masses were added to the pulley. Specifically, the peak lumbar erector spinae muscle activity is reduced and occurs earlier in the flexion movement as the L4-L5 moment is reduced by the counterweights.

Figure 4.3: Average lumbar flexion and trunk inclination angles, L4-L5 moments, and lumbar erector spinae activity of all time-normalized trials. Vertical lines denote the end of lumbar flexion movement and beginning of lumbar extension used to time-normalize trials. During trunk flexion, lumbar erector spinae activity is reduced as greater masses are added to the pulley. Note that due to the experimental set-up the pulley only became engaged at ~20° of lumbar flexion. Note also that during trunk extension, hip extension occurs before lumbar extension, thereby causing trunk inclination to initiate prior to lumbar spine movement.

4.4.2 Abdominal Muscles

Rectus abdominus and external oblique muscles had distinct activation onsets near full flexion. There were no temporal differences between lumbar erector spinae inactivation and abdominal muscle activation for 0, 2.27, and 4.54 kg conditions (Figure 4.4). In the 6.80 kg condition, abdominal muscles activated 0.99

± 0.39 and 1.29 ± 0.42 s before lumbar erector spinae inactivation (p = 0.0462 and p

= 0.0383, respectively).

Rectus abdominus and external oblique onsets occurred with lesser L4-L5 moments when greater masses were attached to the pulley (both p < 0.0001).

Rectus abdominus onset also occurred with less lumbar flexion (p = 0.0112) and less trunk inclination (p =0.0278; Figure 4.5) when greater masses were attached to the pulley. However, this was not the case for external oblique onset with respect to lumbar flexion and trunk inclination (p = 0.1418, and p = 0.2131, respectively;

Figure 4.6). In general, peak abdominal muscle activity was greater when more mass was attached to the pulley (Figure 4.7).

Figure 4.7: Average time-normalized muscle activity of all participants through trunk flexion and return to stand. Note that lumbar erector spinae activity is shown in Figure 4.3.

4.4.3 Additional Muscles

Multifidus, thoracic erector spinae, latissimus dorsi, and internal oblique muscles did not have distinct muscle onsets or offsets. Therefore, time-normalized muscle activity was compared qualitatively. There were no noticeable differences between activity of multifidus at L1 and L4, so these sites were pooled for further analysis. Multifidus activity was generally similar to lumbar erector spinae activity with some differences (Figure 4.7). Rather than being continuously active

throughout the flexion and extension phases of the movement, multifidus demonstrated short bursts of activity separated by periods of inactivity. Multifidus was most often active at the initiation and termination of movements and least active during the identified ‘quiet period’ of lumbar erector spinae activity; however, 5 of 10 participants demonstrated some activity during this period.

Thoracic erector spinae was always partially active (Figure 4.7); however, the change in magnitude throughout the flexion and extension movement was highly variable between participants. In general, thoracic erector spinae activity increased from ~5% to 7.6% MVC during the initial part of trunk flexion and was maintained through the static hold in full flexion. Thoracic erector spinae was most active during the return to standing portion of the movement, which is similar to the activation profile of lumbar erector spinae. When 2.27 and 4.54 kg masses were attached to the pulley thoracic erector spinae activity was reduced during trunk flexion and static hold compared to the 0 kg condition; however, this reduction in activity was not present in the 6.80 kg condition.

Latissimus dorsi muscle activity was less variable between participants than thoracic erector spinae muscle activity. Latissimus dorsi was minimally active in a neutral posture and increased as the participant flexed forward (Figure 4.7).

Latissimus dorsi was most active at full flexion and throughout the static hold. Peak latissimus dorsi activity was reduced in conditions where mass was attached to the pulley compared to the 0 kg condition.

Similar to rectus abdominus and external oblique, internal oblique activity dramatically increased prior to full flexion; internal oblique also activated earlier

and to a larger magnitude when more mass was attached to the pulley (Figure 4.7).

However, unlike rectus abdominus and external oblique, internal oblique did not demonstrate quiet steady-state activity in the neutral standing position. During the initial part of trunk flexion internal oblique activity very slightly decreased; this decrease, followed by a large increase in activation, made it difficult to accurately and consistently identify an onset of muscle activation.

4.5 Discussion

The results of this study support the hypothesis that flexion relaxation of lumbar erector spinae occurs earlier when the required L4-L5 extensor moment is reduced. Less lumbar flexion and consequently less strain of passive lumbar tissues was required for lumbar erector spinae to cease active force production. These results agree with previous studies demonstrating the opposite effect, that increasing the L4-L5 moment caused a delay in lumbar flexion relaxation (Donish and Basmajian, 1972; Schultz et al., 1985; McGill and Kippers, 1994). Thus, this study provides further support for the concept that flexion relaxation occurs in a healthy individual when active force from the lumbar erector spinae is no longer required to equilibrate the external moment. Flexion relaxation therefore has the potential to be a useful research tool for investigating passive spine mechanics in vivo. Because flexion relaxation occurs when the passive tissue moment equals the external moment, manipulating the external moment and identifying when flexion relaxation occurs could determine the relationship between passive spine moments and lumbar flexion. The secondary purpose of this study was to characterize the

activity of the surrounding musculature, and determine the influence of the L4-L5 moment on these muscles. Only multifidus demonstrates flexion relaxation similar to lumbar erector spinae; thoracic erector spinae and latissimus dorsi muscles maintain activity throughout the static hold at full flexion. The abdominal muscles turn ‘on’ at approximately the same time as lumbar erector spinae turn ‘off’.

Decreasing the L4-L5 moment in the 2.27 and 4.54 kg conditions decreased extensor muscle activity and increased abdominal muscle activity. However, further decreasing the moment in the 6.80 kg condition altered the activation timing of abdominal and extensor muscles leading to co-contraction.

Howarth and Mastragostino (2013) added ~8 and 16 kg masses to the trunks of healthy male participants and reported that both L4-L5 moments and lumbar flexion angles increased at the critical point for lumbar erector spinae inactivation.

Combining their results with the results of this study provides clear evidence that the lumbar flexion angle at the instant the lumbar erector spinae cease activity is strongly related to the L4-L5 moment (Figure 4.8). While the relationship between lumbar flexion and L4-L5 moments agree, direct comparisons of lumbar flexion angles between the two studies are difficult as Howarth and Mastragostino (2013) reported angles between T9 and S1, while in the current study angles were calculated between T12 and S1. The current experimental set-up required the pulley to be attached to the thoracic spine making it impossible to attach a rigid body to T9.

Figure 4.8: The relationship between lumbar flexion angle and L4-L5 moment when adding (Howarth and Mastragostino, 2013) or subtracting (current study) mass from the torso.

An important consideration of this study is that the differences in lumbar flexion angle at the critical point of lumbar erector spinae inactivation, despite statistical significance and agreement with the literature, are quite small. Cohen’s D effect sizes (difference of means/pooled standard deviation) comparing lumbar flexion angle at the critical point ranged between 0.10 and 0.31; therefore this small effect size can be interpreted as a real effect that can only be seen through careful observation. More convincing evidence that altering the L4-L5 moment reduces lumbar erector spinae activity can be seen in the time-normalized trace (Figure 4.3); adding mass to the pulley reduced the activity of lumbar erector spinae throughout trunk flexion. Colloca and Hinrichs (2005) have emphasized that discussing lumbar erector spinae dichotomously as either active or inactive may be an oversimplification. This study further supports that the critical point for inactivation

may not be the most sensitive measure of flexion relaxation. Future studies should also consider lumbar erector spinae activity prior to complete muscle inactivation, similar to the time-normalized trace (Figure 4.3), in order to determine changes in activity throughout trunk flexion, rather than simply at the critical point of inactivation.

Contrary to the L4-L5 moment and flexion angle, our hypothesis that trunk inclination angle would decrease at the critical point was refuted. Trunk inclination was not significantly affected by the reduced L4-L5 moment. This may be a limitation of using the counterweights, as the ratio of spine to hip flexion was altered in the 2.27, 4.54, and 6.80 kg trials compared to the 0 kg condition. Further investigation of the spine and hip flexion angles throughout the trials revealed that the pulley had very little effect on peak spine flexion; however, peak hip flexion was

~5° greater in all trials using the pulley.

The results of this study indicate that during trunk flexion the abdominal muscles become active at approximately the same time that lumbar erector spinae becomes inactive, supporting the hypotheses of Solomonow (2003) and Olson et al.

(2004). Decreasing the L4-L5 moment causes both lumbar erector spinae inactivation and abdominal activation to occur earlier, with no statistical difference in timing between the critical points of lumbar erector spinae and abdominal muscles. In the 6.80 kg condition, the synchronization of lumbar erector spinae inactivation and abdominal activation disappeared. In this condition, abdominal muscles activated earlier resulting in a period of increased trunk co-contraction.

This increased co-contraction may be due to the increased challenge in completing

this task, as participants also stated anecdotally that this condition was the most awkward to perform. It is likely that adding additional counterweights to the torso would further increase this co-contraction; however, this remains to be determined.

This is the first study of which we are aware that compares offset/onsets times of lumbar erector spinae and abdominal muscles during flexion relaxation.

Multifidus activity at both L1 and L4 sites was consistent with the findings of

Morris et al., (1962) and Pauly (1966). During trunk flexion, the static hold at full flexion, and subsequent trunk extension to standing, multifidus typically demonstrated short bursts of activity followed by periods of inactivity. It is most frequently active at the initiation and termination of both spine flexion and extension. Multifidus appears to display flexion relaxation comparable to lumbar erector spinae as it is more likely to be active during trunk flexion and extension than during the static hold at full flexion; however, half of the participants also had short bursts of multifidus activity during the identified ‘quiet period’ of flexion relaxation, as similarly reported by Donish and Basmajian (1972).

Neither thoracic erector spinae nor latissimus dorsi demonstrated complete inactivation during trunk flexion. As both muscles cross the L4-L5 intervertebral joint, they actively produce lumbar extension moments; thus even during full cessation of lumbar erector spinae activity, the lumbar extensor moment is not born entirely by passive tissues. In the 0 kg condition, thoracic erector spinae and latissimus dorsi activity remained relatively constant from peak flexion to return to upright. Three participants had reduced thoracic erector spinae activity during the static hold compared to peak flexion, which is consistent with results from Donish &

Basmajian (1972), where only half of their participants demonstrated decreased thoracic erector spinae activity at full flexion. Interestingly, when the moment was reduced in the 2.27 and 4.54 kg conditions, thoracic erector spinae and latissimus dorsi activity decreased prior to maximum flexion. Therefore, thoracic erector spinae and latissimus dorsi begin to show signs of flexion relaxation when the moment is reduced; although these muscles do not approach the level of inactivation demonstrated by lumbar erector spinae. Further reducing the L4-L5 moment in the 6.80 kg condition did not lead to further reductions in thoracic erector spinae and latissimus dorsi activity during the static hold at full flexion. This is likely due to the altered activation timing and increased co-contraction between the abdominal and extensor muscles. Specifically, abdominal muscles activated earlier requiring greater thoracic erector spinae and latissimus dorsi activity than during the 2.27 and 4.54 kg conditions. To our knowledge, this is the first study to record EMG of latissimus dorsi during flexion relaxation.

Flexion relaxation of lumbar erector spinae is highly dependent on the L4-L5 moment; decreasing the L4-L5 moment facilitates flexion relaxation to occur earlier while increasing the L4-L5 moment delays flexion relaxation in a healthy population. It is important to note that the differences in lumbar erector spinae critical point are small but reliable; flexion relaxation occurs 3.2° earlier when the lumbar moment is reduced by ~30 Nm. This is because flexion relaxation occurs in a very stiff region of spine range of motion. In this region, small changes in flexion angle lead to large changes in passive resistance of passive muscles and ligaments.

Conditions that increase or decrease the passive stiffness of spine muscles or

ligaments would be expected to affect the presentation of flexion relaxation.

Although comparing the time normalized trace of lumbar erector spinae activity throughout trunk flexion may be more sensitive to these changes.

While multifidus also exhibits flexion relaxation similar to lumbar erector spinae, the thoracic erector spinae and latissimus dorsi do not demonstrate flexion relaxation. In the natural 0 kg condition, as extensor muscles turn off the abdominal muscles turn on to produce the final degrees of trunk flexion. Coordination between abdominal and extensor muscles can be altered by significantly decreasing the L4-

L5 moment (6.80 kg condition), resulting in a period of increased co-contraction. It is important to recognize that in a healthy population the external moment is a key determinant of lumbar flexion relaxation. As demonstrated here, the increased difficulty of the 6.80 kg condition led to altered activation strategies of abdominal and thoracic muscles. As such, alterations in either the mechanics or activation of these muscles may affect the presentation of flexion relaxation by influencing the lumbar moment. Activation strategies and mechanics of abdominal and thoracic muscles may explain why some clinical populations do not exhibit flexion relaxation; future studies of the mechanism or presentation of flexion relaxation may benefit by including these muscles in their scope.

4.6 Bridge Summary

The primary purpose of this study was to further test the proposed mechanism causing flexion relaxation. Flexion relaxation of lumbar erector spinae is believed to occur because passive tissues become strained and generate extensor

moments. These passive moments are large enough that previously active muscles are no longer required to maintain stability and can ‘turn off’. If this mechanism is true, then decreasing the extensor moment should allow flexion relaxation to occur earlier, as less strain would be needed to balance the external moment. The findings of this study supported this hypothesis; flexion relaxation occurred earlier when the pulley offloaded weight from the torso. The contribution of passive muscle and ligaments to the passive moment responsible for flexion relaxation is unknown. The kinematic and kinetic data from this experiment, along with mechanical and physiological data from Experiments I and II, were used as input to a custom biomechanical model of the spine to evaluate whether flexion relaxation was predicted to occur and whether passive muscles or ligaments contributed to flexion relaxation. The development and results of the musculoskeletal model are presented in Part B.

Part B: Chapter 5 Musculoskeletal Model

5.1 Chapter Summary

Musculoskeletal models are highly valuable research tools for gaining insight into biological phenomena that cannot be tested experimentally. We do not currently have the ability to non-invasively record in vivo forces within muscles and ligaments. However, by combining detailed information of the physiological, mechanical, and architectural properties of muscles and ligaments, computational models can estimate forces within these structures. The purpose of this chapter was to develop a detailed musculoskeletal model of the spine capable to estimating the passive muscle and ligament contributions to the lumped passive spine moment; this model would then be used to determine the relative contributions of muscles and ligaments at the instant of flexion relaxation. There were two steps to the generation and validation of this model.

First, mechanical properties of rat spine muscles (Experiment I) were combined with physiological sarcomere lengths of cadaveric human spine muscles

(Experiment II) and existing architectural data from the literature in order to estimate passive muscle stresses and L4-L5 sagittal moments throughout the physiological range of sagittal plane spine motion. The predicted passive muscle stresses and moments were compared with three previously published passive muscle models, which have been incorporated into musculoskeletal spine models.

The goal of this section was to determine whether predicted muscles stresses and moments derived experimentally from cadaveric and animal spine muscles agree with previous muscle models; this is important, as there is no ‘gold standard’ for 88

passive muscle models. Results of the model demonstrated that predicted passive muscle stresses and moments were similar to those predicted using the model developed by Dunk et al., (2004); both of the models developed by McGill & Norman

(1986) and Thelen (2003) predicted greater muscle stresses, compared to the current model, near end range of lumbar flexion. Overall, the very close agreement between the theoretical model of Dunk and colleagues (2004) and the current experimental model lends credence to the predictions of both models.

Second, the passive muscle model was combined with an existing ligament model to estimate the passive L4-L5 sagittal moments throughout the flexion relaxation movements recorded in Experiment III. The required extensor moments calculated in Experiment III using inverse dynamics were also used as input to the model. These passive muscle, ligament, and required extensor moments were also compared at the critical point of flexion relaxation in order to determine the relative contributions of each tissue. Active muscle was assumed to generate the difference between the required extensor moment and the total passive tissue moment (Figure

5.1). The goal of this section was to determine if the predicted passive muscle and ligament moments were consistent with in vivo experimental data and whether flexion relaxation could be predicted using a spine model. Flexion relaxation is useful for testing the predicted forces of the passive components of the musculoskeletal model, as lumbar muscle are predominantly inactive. In addition, models should be able to predict observed phenomena, such as flexion relaxation.

Model results showed that the predicted active muscle moments were very similar to the activation patterns recorded in Experiment III. Both passive muscle and

ligaments contribute substantially to the lumbar extensor moment at the critical point of flexion relaxation. The results of these models indicate that passive muscle generates larger extensor moments than ligaments near full spine flexion; however, ligaments are more sensitive than passive muscle to small changes in spine flexion at end ranges of motion. The combined passive muscle and ligament moment successfully predicted the ‘flexion relaxation phenomenon’.

Figure 5.1: Overview of input parameters used in the computational spine model. The model predicts active and passive muscle and ligament moments throughout dynamic trunk flexion.

5.2 Passive Muscle Model

5.2.1 Model development

The geometries of the spine and muscles were based off of work by

Cholewicki & McGill (1996). This model represented a 50th percentile male and was composed of seven rigid bodies (thorax, five lumbar vertebrae, and sacrum/pelvis) and 45 bilateral muscle fascicles representing ten spine muscle groups (multifidus,

lumbar erector spinae, iliocostalis thoracis, longissimus thoracis, latissimus dorsi, quadratus lumborum, psoas major, external oblique, internal oblique, and rectus abdominus).

Spine posture was a function of the sagittal lumbar spine angle between the

T12 and S1 rigid bodies. The proportion of lumbar spine flexion that occurred at each motion segment was assigned using population averages (from T12-L1 to L5-

S1: 13, 14, 16, 20, 22, and 15%; McGill & Norman 1986; White & Panjabi 1990;

Potvin et al., 1991; McGill 2007). Rotations were applied sequentially from L5-S1 to

T12-L1.

Muscle fascicles were represented as vectors spanning origin and insertion points attached to the spine (Figure 5.2); some fascicles were forced through nodal points to induce fascicle curvature. Anatomic coordinates, tendon lengths, and muscle physiological cross sectional areas were determined from cadaveric dissections, MRI, and CT images of adult males by McGill and Norman (1986), McGill

(1992), and Bogduk et al., (1992). Fascicle lengths were calculated throughout the physiological range of sagittal spine motion (~60° flexion to 16° extension) using the Euclidean distance between the origin, nodal points (if necessary), and insertion coordinates. Muscle lengths were calculated as the fascicle length minus the tendon length; tendon strain was not included in this model for two reasons: 1) erector spinae aponeurosis and spine muscle tendon stiffness values are not available in the literature, and 2) tendons are significantly stiffer than partially activated muscles

(Kubo et al., 2015). This assumption does not affect multifidus, lumbar erector spinae, quadratus lumborum, or rectus abdominus because these muscles have

negligible tendons. Tendon mechanics for the remaining muscles should be investigated in the future and incorporated into the model.

Figure 5.2: Sagittal view of modelled muscle fascicles of six muscle groups included in the musculoskeletal model.

Physiological neutral spine sarcomere lengths (SLN) from Experiment II were assigned to each muscle fascicle; latissimus and abdominal wall neutral sarcomere lengths were obtained from Gerling & Brown (2013) and Brown et al. (2010), respectively. The sarcomere lengths (SLi) of each muscle were calculated throughout the range of motion using

��! ��! = ∗ ��! ��!

where MLN is the muscle length in the neutral position and MLi is the muscle length in the current model frame (i). Muscle stress was predicted from the sarcomere length using the cubic spline for bundles of multifidus fibres from Experiment I. As passive stress-sarcomere length relationships were unavailable for most muscle groups and there were minimal differences between multifidus and erector spinae of the rat and mouse, passive stresses of all muscles were estimated from multifidus data. Different passive stress-sarcomere length relationships for each muscle could be incorporated in the future when more data are available. Muscle force was calculated as the product of muscle stress and the physiological cross sectional area for each muscle. The L4-L5 sagittal plane moment for each muscle was calculated using the cross-product of the position vector of the L4-L5 rotation centre to the muscle insertion and the muscle force vector; all nodes of curved muscle fascicles were above the L4-L5 motion segment and muscle insertion coordinates could be used for the position vector. The total L4-L5 passive muscle moment was the sum of all individual muscle moments.

5.2.2 Additional passive muscle models

The passive muscle model developed above was compared with three other passive muscle models. Muscle moment arms, fascicle lengths, and physiological cross sectional areas were kept consistent between models in order to isolate the effect of passive muscle stress. McGill and Norman (1986) adapted the work of

Woittiez et al. (1984) on rat calf muscles and modelled passive spine muscles using

! !!".!"#!!.!"# ! �!" = �! ∗ � !

where Fpm is the passive muscle force, PO is the maximum isometric muscle force, LO is the optimum muscle length for force production, and L is the current muscle length. LO was estimated by McGill & Norman (1986) for each muscle assuming that optimal muscle lengths would coincide with a relaxed fetal position. PO was estimated using the product of specific muscle stress (35 N/cm2, McGill & Norman

1986) and physiological cross sectional area.

Dunk and colleagues (2004) found that estimates of passive muscle forces were too large using McGill and Norman’s (1986) model. These authors used the same exponential equation to predict passive muscle forces, however they adjusted the LO of extensor and flexor muscles to occur at 70% of spine flexion and 90% of spine extension, respectively.

Finally, Christophy et al., (2012) incorporated Thelen’s (2003) passive muscle equation into their OpenSim spine model. This model also predicts passive muscle force based on the ratio of muscle length to optimal length using

!! �!.!!! − 1 � = � ∗ !" ! �! − 1

Again, PO was estimated as the product of physiological cross sectional area and specific muscle stress; for this model specific muscle stress was increased to

2 46 N/cm to be consistent with the OpenSim model. In the OpenSim model, LO in based off of physiological sarcomere length measurements from Delp et al., (2001).

These lengths may represent passive muscle slack lengths and not true physiological sarcomere lengths as some measures were obtained from non-fixed muscle tissue dissected from the skeleton which would shorten under it’s own passive tension. For this reason, LO was based off of physiologic sarcomere lengths reported in Experiment II.

5.2.3 Results

5.2.3.1 Passive muscle stress-length relationship

Passive force-length relationships are often depicted by normalizing length and force to optimal muscle length and maximally activated isometric forces, respectively. This could not be done for the current model, as maximally activated isometric forces were not recorded in Experiment I. Rather, passive muscle stresses

(N/cm2) were predicted using the McGill and Norman (1986) and Thelen (2003)

2 models by dividing PO (N) by PCSA (cm )(Figure 5.3). The Dunk et al., (2004) model is not shown, as this model uses the same stress-optimal length relationship as

McGill & Norman (1986). The McGill & Norman (1986) and Dunk et al., (2004) models produce much higher stresses at long normalized muscles lengths compared to the Thelen (2003) or current models.

5.2.3.2 L4-L5 sagittal passive muscle moment

The stresses developed within each muscle depend on the optimal muscle length for active force production, which was estimated differently for each model.

It would be difficult to compare muscle stresses of 90 muscle fascicles of four different models. Instead muscle stresses were converted to L4-L5 moments and summed to obtain the cumulative L4-L5 sagittal passive muscle moment throughout the range of motion for each model (Figure 5.4). Both the Dunk (2004) and current model predicted similar moments throughout the entire range of lumbar motion,

while the McGill (1986) and Thelen (2003) models predicted greater extensor moments near full flexion.

The Dunk et al. (2004) model was the only participant specific muscle model that considered spinal flexibility; the optimal lengths of extensor muscles were set

to occur at 70% of maximum trunk flexion and 90% of maximum trunk extension for flexor muscles. The effect of spine flexibility on predicted passive muscle moments is seen in Figure 5.5 with increasing flexibility shown as the line colors range from yellow to purple. The Dunk et al., (2004) model predicted greater flexor moments compared to the current model throughout the range of motions of less flexible individuals (yellow lines). The current model also predicted passive flexor moments near neutral postures, while the Dunk et al., (2004) model predicted negligible passive moments in neutral postures of all participants.

Incorporating tendon strains into the model will decrease the muscle length, sarcomere length, and passive muscle stresses of the thoracic erector spinae, latissimus dorsi, internal oblique, external oblique, and psoas muscles. The expected magnitudes of tendon strains are difficult to model because the relative stiffness between the muscles and the tendons are unknown. However, if 2% tendon strain is incorporated into the model, the passive muscle moment at maximum flexion decreases from 28.76 Nm to 27.82 Nm.

5.2.4 Implications of these findings

The fact that both the current model and the model developed by Dunk and colleagues (2004) predicted very similar passive muscle moments throughout the range of motion indicates that these models provide realistic estimates of passive muscle stress. Both models converged on similar estimates using very different approaches; Dunk and colleagues intuitively altered an existing passive muscle

model to reflect individual differences in spine flexibility, and the current model is based on experimental data obtained from cadaveric humans and rat spine muscles.

There are two main differences in the predictions of the current model and

Dunk et al., (2004). First, the current model predicted a lesser extension moment near full flexion in less flexible individuals. A limitation of the current model is that it does not depend on spinal flexibility. One way to introduce flexibility into the current model would be to apply a multiplication factor based on range of motion to the cumulative L4-L5 moment; however, both the magnitude of this factor and whether the factor should be applied only at end ranges or throughout the range of motion are unknown. Further, spine ligaments rather than passive muscles may limit full lumbar flexion. While it is agreed that passive muscle moments predicted by spine models should be dependent on spine flexibility, it may not be necessary for both the passive muscle model and the ligament model to be dependent on maximum trunk flexion. It would be beneficial to determine which structures

(ligament, passive or active muscle) limit further spine flexion at end range of motion. Future studies incorporating shear wave elastography may be able to determine this by measuring in vivo muscle stiffness in fully flexed postures. Second, the current model predicted a passive muscle flexion moment in neutral and extended postures, whereas the Dunk et al., (2004) model predicted minimal passive muscle moments. The current model may be more accurate than Dunk et al.,

(2004) in neutral and extended postures, as Brown et al., (2010) recorded physiological sarcomere lengths of abdominal muscles in a neutral posture to be on the descending limb of the active force-length curve. At these long sarcomere

100

lengths, abdominal muscles would be expected to generate passive forces. These passive forces combined with the large abdominal wall moment arms would create substantial passive muscle flexion moments in neutral and extended postures. It should be noted that Dunk et al., (2004) were investigating flexed spine postures; these authors were not attempting to predict passive muscle moments in neutral and extended postures. The Dunk et al., (2004) model could be modified to increase passive flexion moments and predict long sarcomere lengths in neutral postures by setting rest lengths of flexor muscles to occur at 20-30% of full flexion.

Two other important considerations when comparing the current model with the Dunk et al., (2004) model concern the magnitudes of passive muscle stress.

The mechanical properties used in the current model were not obtained from whole muscle testing. Rather, muscle stresses were estimated from bundles of muscle fascicles, which include muscle fibres and extracellular connective tissue. However, it is likely that this connective tissue only includes the basal lamina and endomysium. Perimysium and epimysium surround muscle fascicles and whole muscles, respectively. Including these extracellular structures may increase passive muscles stresses if they have greater moduli than bundles of muscle fibres; however it is unknown to what extent passive stresses would increase. For this reason the current model may underestimate passive muscle stresses at long muscle lengths. If this is the case, then the large extensor moments predicted by the models of McGill

& Norman (1986) and Thelen (2003) may be more accurate at end ranges of motion.

Whole muscle testing will be required to determine the passive stresses at end range of motion. It should also be noted that the passive stresses shown in Figures

101

5.3 and 5.4 do not include the gain factor often employed by musculoskeletal spine models (McGill 1992; Granata & Marras 1995; Cholewicki & McGill 1996; Nussbaum

& Chaffin 1998). Predicted muscle moments are often less than the resultant moment derived from inverse dynamics. Muscle stresses are thus often multiplied by a gain factor of 2-3 in order to balance the resultant moment; however, there is considerable debate whether this gain factor accounts for underestimations of specific muscle stress, physiological cross sectional area, moment arms, or other unknown factors (Reeves & Cholewicki 2003). The passive stresses recorded in

Experiment I were more variable between muscle fibres than they were between muscles, or between animals; since whole muscle stress would be the average stress of a population of muscle fibres (with the addition of perimysium and epimysium) it doesn’t seem likely that at the whole muscle level there would be large differences in passive muscle stress between healthy individuals. Additionally, when P0 in the

Dunk et al., (2004) model is estimated using a specific muscle stress of 35 N/cm2, predicted muscle stresses aligned very closely with the predicted stresses of the current model; specific muscle stress cannot be defined in the current model at it is defined by the maximum stress of activated muscles. As the predicted stresses of the current model and Dunk et al., (2004) model agreed, it is reasonable to conclude that the required gain term does not correct for individual differences in specific muscle stress. It seems that the gain term often employed by EMG driven musculoskeletal models may account for differences in physiological cross sectional area or moment arms between individuals. If this were the case, then a gain term would need to be applied to the current muscle model as well.

102

Finally, both the McGill & Norman (1986) and Thelen (2003) models estimated greater passive muscle extensor moments near full spine flexion. The model used by McGill and Norman (1986) was not developed for full flexion postures and this passive muscle model is no longer used in McGill’s spine model.

Later investigations of full spine flexion noted that these forces were too high

(Potvin et al., 1991); the passive muscle model was replaced with a lumped rotational stiffness parameter, which accounted for muscle, ligaments, lumbodorsal fascia, and skin (Cholewicki & McGill 1996). As the current model estimated the moment specifically due to passive muscle, we could not use the passive lumped rotational stiffness parameter as a comparison. It may also be unrealistic to expect the Thelen (2003) passive muscle model to predict the same stresses as the current model. Christophy et al., (2012) applied the Thelen (2003) passive muscle model to their OpenSim spine model; however this model uses different muscle fascicle geometries, muscle lengths, and physiological cross-sectional areas. As musculoskeletal models are highly sensitive to these parameters it is difficult to combine aspects from various models.

5.3 Model of flexion relaxation

5.3.1 Model development

The ligament model adopted by Potvin et al., (1991) from McGill & Norman

(1986) was added to the spine and passive muscle model used in the previous section (Figure 5.6).

103

The origin and insertion coordinates of 14 ligaments crossing the L4-L5 motion segment are reported by Cholewicki & McGill (1996). Ligament forces are predicted from ligament strain using the exponential function

� = ��!ɛ + �

where α and β are shape fitting coefficients (Anderson et al. 1985), ϕ is a correction for cross-sectional area, ɛ is percent strain (length – rest length)/rest length*100, and P is ligament pre-tension. Variables α, β, ϕ, and P are listed in McGill & Norman

(1986) and McGill & Kippers (1994). Previously, differences in flexibility between

104

individuals were included by setting ligament rest lengths to occur 6° prior to full

L4-L5 flexion (McGill 1988; Potvin et al., 1991). One drawback of this method is that manipulating the rest length does not simply horizontally translate the exponential function; rather it has large effect on the rate of growth. This predicts that ligament forces at maximum flexion are significantly greater in individuals with less range of motion (Figure 5.7). These forces are also much greater than the experimentally determined ligament failure forces reported by Myklebust et al. (1988).

105

In order to ensure maximum ligament forces were not dependent on flexibility and did not exceed the failure forces reported by Myklebust et al., (1988), rest lengths for each ligament were instead evaluated in the current model by assuming that they generated 70% of failure forces at maximum range of motion

(Lmax).

! !"#$!!"#$%"& ∗!"" 0.7�!"# = �� !"#$%"& + �

Using this method, ligament forces were dependent on participant flexibility, yet at end ranges of motion forces were similar between individuals and were physiologically relevant (Figure 5.8).

106

Figure 5.7: Ligament forces of three theoretical participants with different levels of spine flexibility predicted using the current method of evaluating ligament rest lengths. The black dashed line represents the failure force for the supraspinous ligament (Myklebust et al. 1988).

L4-L5 sagittal plane passive muscle and ligament moments were calculated throughout the experimental trunk flexion trials based on the spine kinematics recorded in Experiment III. Passive muscle moments were dependent on lumbar flexion angle and ligament moments were dependent on lumbar flexion and participant flexibility (maximum lumbar flexion angle obtained during any trial).

The passive muscle model was multiplied by a gain factor of 2.0, as this is a representative gain factor often required when predicting muscle moments using the Cholewicki & McGill (1996) model. The difference between the total passive moment (ligament + muscle) and the required external moment calculated using inverse dynamics in Experiment III was assumed to be the moment generated by

107

active muscles. The percentage of the external moment contributed by ligaments, passive and active muscles at the critical point for flexion relaxation was calculated for each condition (0, 2.27, 4.54, and 6.80 kg masses attached to the pulley).

5.3.2 Results of the model

5.3.2.1 Moments throughout trunk flexion movements

Passive muscle produced greater L4-L5 extensor moments than ligaments throughout the normal range of spine motion (Figure 5.9). Slight differences in peak lumbar spine flexion angle between experimental conditions had larger effects on the lumbar ligament moment than on the passive muscle moment due to the greater stiffness of ligaments near end range of motion.

Figure 5.9: Average passive muscle, ligament, and total passive L4-L5 sagittal plane moments for each condition throughout the trunk flexion movement. Larger passive muscle moments in the 0 and 2.27 kg conditions were due small increases in participant flexion angle. The passive muscle moments were less sensitive to these small differences in flexion angle than the ligament moments.

108

The predicted active muscle moment, which was estimated as the difference between the required extensor moment (from inverse dynamics) and the predicted total passive moment (using the model), very closely resembled the lumbar erector spinae activations recorded in Experiment III (Figure 5.10). During neutral standing, active muscles were estimated to generate a net extensor moment of ~12-20 Nm.

During trunk flexion, the predicted active moment increased to ~60 Nm of extension in the control 0 kg condition. Estimates of peak active moments were reduced as the pulley offloaded more weight. Prior to full lumbar flexion, the predicted active muscle extensor moment decreased, requiring minimal extensor activity in full flexion. As the pulley offloaded weight in full trunk flexion, the model predicts that muscles were required to actively generate a trunk flexion moment.

This is because the modelled passive extension moment exceeded the required extensor moment calculated with inverse dynamics.

109

5.3.2.2 Moments at the critical point of flexion relaxation

At the critical point of flexion relaxation, the model predicted passive muscle accounted for 61.7% of the required extensor moment. Ligaments accounted for

36.5% and active muscle contributed 1.8% (Table 5.1). Adding 2.27 and 4.54 kg masses to the pulley reduced the required extensor moment and increased the relative contribution of passive muscle compared to ligaments. Adding 6.80 kg to the pulley did not cause flexion relaxation to occur earlier than in the 4.54 kg condition. This is the only condition where the model predicted active muscles generated lumbar flexor moments, instead of extensor moments, at the critical point of flexion relaxation.

110

Table 5.1: Contribution of ligament, passive muscle, and active muscle tissue to the required L4-L5 extensor moment at the critical point of flexion relaxation.

0 kg 2.27 kg 4.54 kg 6.80 kg

% Full flexion 95.0 % 93.4 % 91.8 % 92.2 %

Required L4-L5 moment (Nm) 91.4 88.6 82.6 77.9

Ligament moment (Nm) 33.3 29.1 24.4 24.9 36.5 % 32.8 % 29.5 % 32.0 %

Passive muscle moment (Nm) 56.4 55.1 53.4 53.7 63.5 % 64.0 % 67.0 % 70.7 %

Active muscle moment (Nm) 1.7 4.4 4.8 -0.7 1.8 % 4.9 % 5.8 % -0.9 % %Full flexion is the lumbar flexion angle at the critical point of flexion relaxation as a percentage of the maximum flexion angle recorded from any trial.

5.3.3 Implications of these findings

The results of this model agreed very closely with the findings of experiment

III. The predicted active muscle moment appeared very similar to the recorded lumbar erector spinae EMG. At the critical point of flexion relaxation the model predicted the total passive tissue moment was within 5 Nm of the required extensor moment. This strongly supports that both the ligament and passive muscle models are predicting realistic moments. Passive muscle contributes a greater proportion of the extensor moment near full flexion; however, the stiffer ligaments are more sensitive to small changes in lumbar angle, as can be seen in Figure 5.9. This explains why the 2.27 and 4.54 kg pulley conditions had large effects on the required extensor moment, but relatively little change in the lumbar flexion angle at the critical point for flexion relaxation. It seems that near full flexion, even though

111

passive muscle is supporting a larger percentage of the extensor moment, the critical point for flexion relaxation is largely influenced by the stiffness of the spinal ligaments.

The results of the 6.80 kg condition were very interesting. Even though the

6.80 kg condition reduced the required L4-L5 extensor moment, flexion relaxation did not occur earlier than the 4.54 kg condition. This was the only condition that the model predicted the active muscles to generate a flexor moment at the critical point of flexion relaxation. This agrees with the observation in Experiment III that abdominal muscles turned on before lumbar erector spinae muscles turned off in this condition. This provides further support that abdominal muscle activity is able to modulate the presentation of flexion relaxation. Patients with low-back pain often maintain lumbar erector spinae activity in full flexion; this could be due to increased abdominal muscle activation in this posture. To the best of my knowledge, no study investigating flexion relaxation in low back pain patients has recorded abdominal muscle activity. This increased co-activation may be a method for increasing joint stiffness in order to maintain stability (Cholewicki & McGill 1996); however it would also lead to increased compression and shear forces on the intervertebral discs.

Further research is warranted to determine whether the absence of flexion relaxation is an adaptive or maladaptive response to low-back pain.

It was also observed in Experiment III that lumbar erector spinae and multifidus were the only muscles to exhibit flexion relaxation. While thoracic erector spinae activity decreased near full flexion, it was still actively producing force; latissimus dorsi, rectus abdominus, external oblique, and internal oblique

112

were also active at full flexion. This could be due to a requirement to simultaneously maintain spine stability at all lumbar segments. While passive muscle forces are sufficient to stabilize the L4-L5 motion segment, abdominal and thoracic muscle activity may be required in order to stabilize superior or inferior motion segments.

As the total active muscle moment at the critical point of flexion relaxation was predicted to be less than 5 Nm, it appears that the active extensor moments generated by thoracic erector spinae and latissimus dorsi are equal to the flexor moments generated by the abdominal muscles.

113

Chapter 6 Conclusions and future directions

The focus of this thesis was to enhance the understanding of passive and active properties of spine muscles. This work has demonstrated the complexity of muscles, inherent differences between them, and the need for further investigation.

By studying the mechanical properties, physiological design, and activation patterns of multiple lumbar muscles, this thesis identified a number of novel findings:

• Partial tenotomy of the erector spinae aponeurosis does not alter passive

muscle stresses after 28 days.

• Spine muscles are physiologically designed to generate the greatest active

forces in a mid-flexed posture.

• Flexion relaxation can be manipulated to occur earlier by decreasing the

weight of the trunk.

• Abdominal muscles actively generate flexor moments at full trunk flexion

• Increased co-contraction can delay flexion relaxation.

• Passive muscles and ligaments contribute to the flexion relaxation

phenomenon.

Muscles have a large influence on spine loading, stiffness, and stability. In

order to improve model estimates, inherent differences between muscles need

to be further understood. While this thesis has made significant contributions to

the understanding of passive and active muscle properties, there are still many

improvements that can be made to computational models of the spine.

114

This model has only investigated the lumbar L4-L5 sagittal plane moment.

Most spine models focus on either the L4-L5 (McGill & Norman 1986; McGill

1988; 1992) or L5-S1 (Granata & Marras 1995) motion segments as these

vertebrae and discs are the most common sites of spine injury. I have expanded

the current model to investigate forces and moments at the remaining motion

segments as well as during axial twist and lateral bend movements; however,

this work is outside the scope of this thesis. It is important to understand that

almost all lumbar spine muscle fascicles act across multiple motion segments.

This has large implications for maintaining static and dynamic equilibrium.

Muscles cannot simply produce active and passive forces in order to stabilize a

specific motion segment; rather all motion segments need to be stable at the

same time. Spine motor control is an emerging field in biomechanics and is

focused on determining how specific muscle activation patterns are selected.

Future goals of this model will be to incorporate a non-linear optimization

algorithm to partition the active muscle moment among the 90 muscle fascicles

represented. These data can then be used to estimate spine stiffness, stability,

and loading and this model can be used to study the motor control of the spine.

115

References

Adams, M.A., & Hutton, W.C. (1982). Prolapsed intervertebral disc. A hyperflexion injury 1981 Volvo award in basic science. Spine 7, 184–191.

Ahern, D.K., Follick, M.J., Council, J., Laser-Wolston, N., & Litchman, H. (1988). Comparison of lumbar paravertebral EMG patterns in chronic low back pain patients and non-patient controls. Pain 34(2), 153-160.

Allen, C.E.L. (1948). Muscle action potentials used in the study of dynamic anatomy. The British Journal of Physical Medicine 11(3), 66-73.

Andersson, G.B., Ortengren, R., & Nachemson, A. (1977). Intradiskal pressure, intra- abdominal pressure and myoelectric back muscle activity related to posture and loading. Clinical Orthopaedics and Related Research 129, 156–164.

Anderson, C.K., Chaffin, D.B., & Herrin, G.D. (1985). A biomechanical model of the lumbosacral joint during lifting activities. Journal of Biomechanics 18(8), 571- 584.

Andersson, E.A., Oddsson, L.I.E., Grundström, H., Nilsson, J., & Thorstensson, A. (1996). EMG activities of the quadratus lumborum and erector spinae muscles during flexion-relaxation and other motor tasks. Clinical Biomechanics 11(7), 392-400.

Baker, J.H., & Hall-Craggs, E.C. (1978). Changes in length of sarcomeres following tenotomy of the rat soleus muscle. The Anatomical Record 192(1), 55-58.

Baker, J.H., & Hall-Craggs, E.C. (1980). Changes in sarcomere length following tenotomy in the rat. Muscle & Nerve 3, 413-416.

Beaudette, S.M., Unni, R., & Brown, S.H.M. (2014). Electromyographic assessment of isometric and dynamic activation characteristics of the latissimus dorsi muscle. Journal of Electromyography and Kinesiology 24(3), 430-436.

Bergmark, A. (1989). Stability of the lumbar spine. Acta Orthopaedica Scandinavica Supplementum 230(60), 1-54.

Bogduk, N. (1980). A reappraisal of the anatomy of the human lumbar erector spinae. Journal of Anatomy 131, 525–540.

Bogduk, N., Pearcy, M., & Hadfield, G. (1992) Anatomy and biomechanics of psoas major. Clinical Biomechanics 7(2), 109-119.

Bogduk, N. (2005). Clinical anatomy of the lumbar spine and sacrum. 4th ed. Churchill Livingstone. 116

Brown, S.H.M., & McGill, S.M. (2005). Muscle force-stiffness characteristics influence joint stability: A spine example. Clinical Biomechanics 20(9), 917-922.

Brown, S.H.M., & Potvin, J.R. (2005). Constraining spine stability levels in an optimization model leads to the prediction of trunk muscle cocontraction and improved spine compression force estimates. Journal of Biomechanics 38(4), 745-754.

Brown, S.H.M., & Potvin, J.R. (2007). Exploring the geometric and mechanical characteristics of the spine musculature to provide rotational stiffness to two spine joints in the neutral posture. Human Movement Sciences 26(1), 113-23.

Brown, S.H.M., & McGill, S.M. (2009). Transmission of muscularly generated force and stiffness between layers of the rat abdominal wall. Spine 34(2), E70-E75.

Brown, S.H.M., & McGill, S.M. (2010). A comparison of ultrasound and electromyography measures of force and activation to examine the mechanics of abdominal wall contraction. Clinical Biomechanics 25(2), 115-123.

Brown, S.H.M. (2011). Mechanically relevant consequences of the composite laminate-like design of the abdominal wall muscles and connective tissues. Medical Engineering and Physics 34(4), 521-523.

Brown, S.H.M., Gregory, D.E., Carr, J.A., Ward, S.R., Masuda, K., & Lieber, R.L. (2011). ISSLS Prize Winner: Adaptions to the multifidus muscle in response to experimentally induced intervertebral disc degeneration. Spine 36(21), 1729- 1736.

Chaffin, D.B. (1969) A computerized biomechanical model-development of and use in studying gross body actions. Journal of Biomechanics 2(4),429-441.

Cholewicki, J., & McGill, S.M. (1992). Lumbar posterior ligament involvement during extremely heavy lifts estimated from fluoroscopic measurements. Journal of Biomechanics 25(1), 17-28.

Cholewicki, J., & McGill, S.M. (1996). Mechanical stability of the in vivo lumbar spine: implications for injury and chronic low back pain. Clinical Biomechanics 11, 1– 15.

Christophy, M., Senen, N.A.F., Lotz, J.C., & O'Reilly, O.M. (2012). A musculoskeletal model for the lumbar spine. Biomechanics and Modeling in Mechanobiology 11(1-2), 19-34.

Colloca, C.J., & Hinrichs, R.N. (2005). The biomechanical and clinical significance of the lumbar erector spinae flexion-relaxation phenomenon: A review of 117

literature. Journal of Manipulative and Physiological Therapeutics 28(8), 623- 631.

Crisco, J.J., Panjabi, M.M., Yamamoto, I., & Oxland, T.R. (1992). Euler stability of the human ligamentous lumbar spine. Part II: Experiment. Clinical Biomechanics 7(1), 27-32.

Daggfeldt, K., & Thorstensson, A. (2003). The mechanics of back-extensor torque production about the lumbar spine. Journal of Biomechanics 36(6), 815-825.

Danneels, L.A., Coorevits, P.L., Cools, A.M., Vanderstraeten, G.G., Cambier, D.C., Witvrouw, E.E., & De Cuyper, H.J. (2002). Differences in electromyographic activity in the multifidus muscle and the iliocostalis lumborum between healthy subjects and patients with sub-acute and chronic low back pain. European Spine Journal 11, 13–19. de Bruin, M., Smeulders, M.J., Kreulen, M., Huijing, P.A., & Jaspers, R.T. (2014). Intramuscular connective tissue differences in spastic and control muscle: A mechanical and histological study. PLoS ONE 9(6), 1-11.

Delp, S.L., Suryanarayanan, S., Murray, W.M., Uhlir, J., & Triolo, R.J. (2001). Architecture of the rectus abdominis, quadratus lumborum, and erector spinae. Journal of Biomechanics 34, 371–375.

Dickey, J.P., McNorton, S., & Potvin, J.R. (2003). Repeated spinal flexion modulates the flexion-relaxation phenomenon. Clinical Biomechanics 18(9), 783-9.

Donisch, E.W., & Basmajian, J.V. (1972). Electromyography of deep back muscles in man. The American Journal of Anatomy 133(1), 25-36.

Dufour, J.S., Marras, W.S., & Knapik, G.G. (2013). An EMG-assisted model calibration technique that does not require MVCs. Journal of Electromyography and Kinesiology 23, 608-613.

Dunk, N.M., Brown, S.H.M., & Callaghan, J.P. (2004) Distribution of tissue loads supporting the low back during the flexion relaxation phenomenon in sitting and standing postures. Canadian Society for Biomechanics Conference Proceedings.

Edman, K.A., Elzinga, G., & Noble, M.I. (1978). Enhancement of mechanical performance by stretch during tetanic contractions of vertebrate skeletal muscle fibres. The Journal of Physiology 281, 139-155.

Fan, S., Hu, Z., Zhao, F., Zhao, X., Huang, Y., & Fang, X. (2010). Multifidus muscle changes and clinical effects of one-level posterior lumbar interbody fusion:

118

minimally invasive procedure versus conventional open approach. European Spine Journal 19(2), 316-324.

Floyd, W.F., & Silver, P.H.S. (1951). Function of erectores spinae in flexion of the trunk. Lancet 1(6647), 133-134.

Floyd, W.F., & Silver, P.H.S. (1955). The function of the erectores spinae muscles in certain movements and postures in man. The Journal of Physiology 129, 184– 203.

Fortuna, R., Vaz, M.A., Sawatsky, A., Hart, D.A., & Herzog, W. (2014). A clinically relevant BTX-A injection protocol leads to persistent weakness, contractile material loss, and an altered mRNA expression phenotype in rabbit quadriceps muscles. Journal of Biomechanics 48(10), 1700-1706.

Fridén, J., & Liever, R.L. (2003). Spastic muscle cells are shorter and stiffer than normal cells. Muscle & Nerve 27(2), 157-164.

Gardner-Morse, M., Stokes, I.A., & Laible, J.P. (1995). Role of muscles in lumbar spine stability in maximum extension efforts. Journal of Orthopaedic Research 13(5):802-808.

Geisser, M.E., Ranavaya, M., Haig, A.J., Roth, R.S., Zucker, R., Ambroz, C., & Carusa, M. (2005). A meta-analytic review of surface electromyography among persons with low back pain and normal healthy controls. The Journal of Pain 6(11), 711- 726.

Gerling, M.E., & Brown, S.H.M. (2013). Architectural analysis and prediced functional capability of the human latissimus dorsi muscle. Journal of Anatomy 223(2), 112-122.

Gillies, A.R., & Lieber, R.L. (2011). Structure and function of the skeletal muscle extracellular matrix. Muscle Nerve 44, 318-331.

Glass, D.J. (2005). Skeletal muscle hypertrophy and atrophy signaling pathways. The International Journal of Biochemistry & Cell Biology 37(10), 1974-1984.

Golding, J.S.R. (1952). Electromyography of the erector spinae in low back pain. Postgraduate Medical Journal 28(321), 401-406.

Gollapudi, S.K., & Lin, D.C. (2009). Experimental determination of sarcomere force- length relationship in type-1 human skeletal muscle fibres. Journal of Biomechanics 42, 2011-2016.

Gordon, A.M., Huxley, A.F., & Julian, F.J. (1966). The variation in isometric tension with sarcomere length in vertebrate muscle fibres. The Journal of Physiology 119

184, 170–192.

Granata, K.P., & Marras, W.S. (1995). An EMG-assisted model of trunk loading during free-dynamic lifting. Journal of Biomechanics 28(11), 1309-1317.

Gray, H. (1985). Gray's Anatomy: Fifteenth Edition. Chancellor Press, Great Britain.

Gsell, K.Y. (2016). Multifidus and erector spinae muscle fibre passive stiffness is lower in conjunction with greater lumbar spine neutral zone stiffness in ENT1- deficient mice. M.Sc. Thesis.

Gunning, J.L., Callaghan, J.P., & McGill, S.M. (2001). Spinal posture and prior loading history modulate compressive strength and type of failure in the spine: a biomechanical study using a porcine cervical spine model. Clinical Biomechanics 16, 471–480.

Gupta A. (2001). Analyses of myo-electrical silence of erectors spine. Journal of Biomechanics 34(4),491-6.

Hashemirad, F., Talebian, S., Hatef, B., & Kahlaee, A.H. (2009). The relationship between flexibility and EMG activity pattern of the erector spinae muscles during trunk flexion-extension. Journal of Electromyography and Kinesiology 19(5), 746-753.

Henry, J.L., Yashpal, K., Vernon, H., Kim, J., & Im, H.J. (2012). Lumbar facet joint compressive injury induces lasting changes in local structure, nociceptive scores, and inflammatory mediators in a novel rat model. Pain Research and Treatment 2012(4), 1-11.

Herzog, W., & Leonard, T.R. (1997) Depression of cat solues-forces following isokinetic shortening. Journal of Biomechanics 30(9), 865-872.

Hill, A.V. (1938). The heat of shortening and the dynamic constants of muscle. Proceedings of the Royal Society of London Series B, Biological Sciences 126(843), 136-195.

Howarth, S.J., & Mastragostino, P. (2013). Use of kinetic and kinematic data to evaluate load transfer as a mechanism for flexion relaxation in the lumbar spine. Journal of Biomechanical Engineering 135(10), 1-6.

Hu, B., Ning, X., & Nimbarte, A.D. (2013). The changes of lumbar muscle flexion- relaxation response due to laterally slanted ground surfaces. Ergonomics 56(8), 1295-1303.

Hu, Z.J., Zhang, J.F., Xu, W.B., Zhao, F.D., Wang, J.Y., Fan, S.W., & Fang, X.Q. (2015). Effect of pure muscle retraction on multifidus injury and atrophy after posterior 120

lumbar spine surgery with 24 weeks observation in a rabbit model. European Spine Journal.

Huxley, H., & Hanson, J. (1954). Changes in the cross-striations of muscle during contraction and stretch and their structural interpretation. Nature 173(4412), 973-976.

Huxley, A.F., & Niedergerke, R. (1954). Structural changes in muscle during contraction; interference microscopy of living muscle fibres. Nature 173(4412), 971-973.

Hyun, S.J., Kim, Y.B., Kim, Y.S, Park, S.W., Nam, T.K., Hong, H.J., & Kwon, J.T. (2007). Postoperative changes in paraspinal muscle volume: comparison between paramedian interfascial and midline approaches for lumbar fusion. Journal of Korean Medical Science 22(4), 646-651.

Jamali, A.A., Afshar, P., Abrams, R.A., & Lieber, R.L. (2000). Skeletal muscle response to tenotomy. Muscle Nerve 23(6), 851-862.

Jin, S., Ning, X., & Mirka, G.A. (2012). An algorithm for defining the onset and cessation of the flexion-relaxation phenomenon in the low back musculature. Journal of Electromyography and Kinesiology 22(3), 376-382.

Jósza, L., Kannus, P., Thöring, J., Reffy, A., Järvinen, M., & Kvist, M. (1990). The effect of tenotomy and immobilisation on intramuscular connective tissue. A morphometric and microscopic study in rat calf muscles. The Journal of Bone and Joint Surgery 72-B(2), 293-297.

Kavcic, N., Grenier, S., & McGill, S.M. (2004). Quantifying tissue loads and spine stability while performing commonly prescribed low back stabilization exercises. Spine 29(20), 2319-2329.

Kawaguchi, Y., Yabuki, S., Styf, J., Olmarker, K., Rydevik, B., Matsui, H., & Tsuji, H. (1996). Back muscle injury after posterior lumbar spine surgery. Topographic evaluation of intramuscular pressure and blood flow in the porcine back muscle during surgery. Spine 21(22), 2683-2688.

Kawaguchi, Y., Matsui, H., Gejo, R., & Tsuji, H. (1998). Preventive measures of back muscle injury after posterior lumbar spine surgery in rats. Spine 23(21), 2282- 2288.

Kim, K.T., Lee, S.H., Suk, K.S., & Bae, S.C. (2006). The quantitative analysis of tissue injury markers after mini-open lumbar fusion. Spine 31(6), 712-716.

Kippers, V., & Parker, A.W. (1984). Posture related to myoelectric silence of erectors spinae during trunk flexion. Spine 9(7), 740-745. 121

Kubo, K., Miyazaki, D., Yamada, K., Shimoju, S., & Tsunoda N. (2015) Passive and active muscle stiffness in plantar flexors of long distance runners. Journal of Biomechanics 48(10) 1937-1943.

Larivière, C., Arsenault, A.B., Gravel, D., Gagnon, D., & Loisel, P. (2003). Surface electromyography assessment of back muscle intrinsic properties. Journal of Electromyography and Kinesiology 13, 305-318.

Lieber, R.L., Fazeli, B.M., & Botte, M.J. (1990). Architecture of selected wrist flexor and extensor muscles. The Journal of Hand Surgery 15, 244–250.

Lieber, R.L., & Fridén, J. (2000). Functional and clinical significance of skeletal muscle architecture. Muscle & Nerve 23, 1647–1666.

Lieber, R.L., Loren, G.J., & Fridén, J. (1994). In vivo measurement of human wrist extensor muscle sarcomere length changes. Journal of Neurophysiology 71, 874– 881.

Linke, W.A., Popov, V.I., & Pollack, G.H. (1994) Passive and active tension in single cardiac myofibrils. Biophysics Journal 67, 782-792.

Macintosh, J.E., Valencia, F., Bogduk, N., & Munro, R.R. (1986). The morphology of the human lumbar multifidus. Clinical Biomechanics 1(4), 196-204.

Macintosh, J.E., & Bogduk, N. (1986). The biomechanics of the lumbar multifidus. Clinical Biomechanics 1(4), 205-213.

Mannion, A.F., Taimela, S., Müntener, M., & Dvorak, J. (2001). Active therapy for chronic low back pain; effects of back muscle activation, fatigability, and strength. Spine 26(8), 897-908.

Marras, W.S., King, A.I., & Joynt, R.L. (1984). Measurements of loads on the lumbar spine under isometric and isokinetic conditions. Spine 9(2), 176-188.

Marras, W.S., & Sommerich, C.M. (1991) A three-dimensional motion model of loads on the lumbar spine: I. Model structure. Human Factors 33(2), 123-137.

Marras, W.S., & Granata, K.P. (1997). Spine loading during trunk lateral bending motions. Journal of Biomechanics 30(7), 697-703.

Matthiasdottir, S., Hahn, M., Yaraskavitch, M., & Herzog, W. (2014). Muscle and fascicle excursion in children with cerebral palsy. Clinical Biomechanics 29(4), 458-462.

McGill, S.M., & Norman, R.W. (1986). Partitioning of the L4-L5 dynamic moment into 122

disc, ligamentous, and muscular components during lifting. Spine 11, 666-678.

McGill, S.M. (1988). Estimation of force and extensor moment contributions of the disc and ligaments at L4-L5. Spine 13(12), 1395-1402.

McGill, S.M. (1991). Kinetic potential of the lumbar trunk musculature about three orthogonal orthopaedic sxes in extreme postures. Spine 16, 809–815.

McGill, S.M. (1992). Kinetic potential of the lumbar trunk musculature about three orthogonal orthopaedic axes in extreme postures. Spine 16, 809-825.

McGill, S.M., & Kippers, V. (1994). Transfer of loads between lumbar tissues during the flexion-relaxation phenomenon. Spine 19, 2190–2196.

McGill, S.M., Juker, D., & Kropf, P. (1996). Appropriately placed surface EMG electrodes reflect deep muscles activity (psoas, quadratus lumborum, abdominal wall) in the lumbar spine. Journal of Biomechanics 29, 1503-1507.

McGill, S.M., Grenier, S., Kavcic, N., & Cholewicki, J. (2003). Coordination of muscle activity to assure stability of the lumbar spine. Journal of Electromyography and Kinesiology 13, 353–359.

McGill, S.M. (2007). Normal and injury mechanics of the lumbar spine. In: Low back disorders: evidenced-based prevention and rehabilitation. Human Kinetics, Windsor ON, pp 72-111.

Moo, E.K., Fortuna, R., Sibole, S., Abusara, Z., and Herzog, W. (2016) Sarcomere lengths and sarcomere elongations are not uniform in an intact muscle. Canadian Society for Biomechanics Annual Meeting, Hamilton ON.

Moore, A., Mannion, J., & Moran, R.W. (2015). The efficacy of surface electromyographic biofeedback assisted stretching for the treatment of chronic low back pain: A case-series. Journal of Bodywork & Movement Therapies 19(1), 8-16.

Moreside, J.M., Vera-Garcia, F.J., & McGill, S.M. (2008). Neuromuscular independence of abdominal wall muscles as demonstrated by middle-eastern style dancers. Journal of Electromyography and Kinesiology 18(4), 527-537.

Morin, F., & Portnoy, H. (1956). Electromyographic study of postural muscles in various positions and movements. American Journal of Physiology 186(1), 122- 126.

Morris, J.M., Lucas, D.B., & Bresler, B. (1961). Role of the trunk in stability of the spine. The Journal of Bone and Joint Surgery (American) 43(4), 327-351.

123

Myklebust, J.B., Pintar, F., Yoganandan, N., Cusick, J.F., Maiman, D., Myers, T.J., & Sances, A. (1988). Tensile strength of spinal ligaments. Spine 13(5), 526-531.

Neblett, R., Mayer, T.G., Gatchel, R.J., Keeley, J., Proctor, T., & Anagnostis, C. (2003). Quantifying the lumbar flexion-relaxation phenomenon. I. Theory, normative data, and clinical applications. Spine 28(13), 1435-1446.

Neblett, R., Mayer, T.G., & Brede, E. (2010). Correcting abnormal flexion-relaxation in chronic lumbar pain: responsiveness to a new biofeedback training protocol. The Clinical Journal of Pain 26, 403-409.

Nussbaum, M.A., & Chaffin, D.B. (1998). Lumbar muscle force estimation using a subject-invariant 5-parameter EMG-based model. Journal of Biomechanics 31, 667-672.

Olson, M., Li, L., & Solomonow, M. (2004). Flexion-relaxation response to cyclic lumbar flexion. Clinical Biomechanics 19(8), 769-776.

Olson, M., Solomonow, M., & Li, L. (2006). Flexion-relaxation response to gravity. Journal of Biomechanics 39(14), 2545-2554.

Pauly, JE. (1966). An electromyographic analysis of certain movements and exercises. I. Some deep muscles of the back. Anatomical Record 155(2), 223-234.

Pitcher, G.M., Ritchie, J., & Henry, J.L. (1999). Paw withdrawal threshold in the von Frey hair test is influenced by the surface on which the rat stands. Journal of Neuroscience Methods 87, 185-193.

Potvin, J.R., McGill, S.M., & Norman, R.W. (1991). Trunk muscle and lumbar ligament contributions to dynamic lifts with varying degrees of trunk flexion. Spine 16(9), 1099-1107.

Potvin, J.R., & Brown, S.H.M. (2004). Less is more: high pass filtering, to remove up to 99% of the surface EMG signal power, improves EMG-based biceps brachii muscle force estimates. Journal of Electromyography and Kinesiology 14(3), 389- 399.

Prado, L. G., Makarenko, I., Andresen C., Krüger, M., Opitz, C.A., & Linke, W.A. (2005). Isoform diversity of giant proteins in relation to passive and active contractile properties of rabbit skeletal muscles. The Journal of General Physiology 126(5), 461-480.

Reeves, N.P., & Cholewicki, J. (2003). Modeling the human lumbar spine for assessing spinal loads, stability, and risk of injury. Critical Reviews in Biomedical Engineering 31(1-2), 73-139.

124

Regev, G.J., Kim, C.W., Tomiya, A., Lee, Y.P., Ghofrani, H., Garfin, S.R., Lieber, R.L., & Ward, S.R. (2011). Psoas muscle architectural design, in vivo sarcomere length range, and passive tensile properties support its role as a lumbar spine stabilizer. Spine 36, E1666–E1674.

Rosatelli, A.L., Ravichandiran, K., & Agur, A.M. (2008). Three-dimensional study of the musculotendinous architecture of lumbar multifidus and its functional implications. Clinical Anatomy 21, 539–546.

Sarti, M.A., Lisón, J.F., Monfort, M., & Fuster, M.A. (2001). Response of the flexion- relaxation phenomenon relative to the lumbar motion to load and speed. Spine 26(18), E421-E426.

Sato, E.J., Killian, M.L., Choi, A.J., Lin, E., Esparza, M.C., Galatz, L.M., … & Ward, S.R. (2014). Skeletal muscle fibrosis and stiffness increase after rotator cuff tendon injury and neuromuscular compromise in a rat model. Journal of Orthopaedic Research 32(9), 1111-1116.

Schinkel-Ivy, A., Nairn, B.C., & Drake J.D.M. (2013). Evaluation of methods for the quantification of the flexion-relaxation in the lumbar erector spinae muscles. Journal of Manipulative and Physiological Therapeutics 36(6), 349-358.

Schinkel-Ivy, A., Nairn, B.C., & Drake, J.D.M. (2014). Quantification of the lumbar flexion-relaxation phenomenon: comparing outcomes of lumbar erector spinae and superficial lumbar multifidus in standing full trunk flexion and slumped sitting postures. Journal of Manipulative and Physiological Therapeutics 37(7), 494-501.

Schultz, A.B., Andersson, G., & Ortengren, R. (1982). Loads on the lumbar spine. Validation of a biomechanical analysis by measurements of intradiscal pressures and myoelectric signals. The Journal of Bone and Joint Surgery 64(5), 713-720.

Schultz, A.B., Haderspeck-Grib, K., Sinkora, G., & Warwick, D.N. (1985). Quantification studies of flexion-relaxation phenomenon in the back muscles. Journal of Orthopaedic Research 3(2), 189-197.

Shah, S.B., & Lieber, R.L. (2003). Simultaneous imaging and functional assessment of cytoskeletal protein connections in passively loaded single muscle cells. Journal of Histochemistry & Cytochemistry 51(1), 19-29.

Shah, S.B., Davis, J., Weisleder, N., Kostavassili, I., McCulloch, A.D., Ralson, E., … & Lieber, R.L. (2004). Structural and functional roles of desmin in mouse skeletal muscle during passive deformation. Biophysical Journal 86(5), 2993-3008.

125

Shin, G., Shu, Y., Li, Z., Jiang, Z., & Mirka, G.A. (2004). Influence of knee angle and individual flexibility on the flexion-relaxation response of the low back musculature. Journal of Electromyography and Kinesiology 14(4), 485-494.

Shin, G., & Mirka, G.A. (2007). An in vivo assessment of the low back response to prolonged flexion: Interplay between active and passive tissues. Clinical Biomechanics 22(9), 965-971.

Shirado, O., Ito, T., Kaneda, K., & Strax, T.E. (1995). Flexion-relaxation phenomenon in the back muscles. A comparative study between healthy subjects and patients with chronic low back pain. American Journal of Physical Medicine and Rehabilitation 74(2), 139-144.

Shirazi-Adl, A., Sadouk, S., Parnianpour, M., Pop, D., & El-Rich, M. (2002). Muscle force evaluation and the role of posture in human lumbar spine under compression. European Spine Journal 11(6), 519-526.

Sihvonen, T., Herno, A., Paljärvi, L., Airaksinen, O., Partanen, J., & Tapaninaho, A. (1993). Local denervation atrophy of paraspinal muscles in postoperative failed back syndrome. Spine 18(5), 575-581.

Smith, L.R., & Barton, E.R. (2014). Collagen content does not alter the passive mechanical properties of fibrotic skeletal muscle in mdx mice. American Journal of Physiology - Cell Physiology 306(10), C889-C898.

Snook, S.H., Webster, B.S., McGorry, R.W., Fogleman, M.T., & McCann, K.B. (1998). The reduction of chronic nonspecific low back pain through the control of early morning lumbar flexion: a randomized controlled trial. Spine 23, 2601–2607.

Solomonow, M., Baratta, R.V., Banks, A., Freudenberger, C., & Zhou, B.H. (2003). Flexion-relaxation response to static lumbar flexion in males and females. Clinical Biomechanics 18(4), 273-279.

Staudenmann, D., Potvin, J.R., Kingma, I., Stegeman, D.F., & van Dieën, J.H. (2007). Effects of EMG processing on biomechanical models of muscle joint systems: Sensitivity of trunk muscle moments, spinal forces, and stability. Journal of Biomechanics 40(4), 900-909.

Stokes, A., & Gardner-Morse, M. (1995). Lumbar spine maximum efforts and muscle recruitment patterns predicted by a model with multijoint muscles and joints with stiffness. Journal of Biomechanics 28(2), 173-186.

Stolov, W.C., & Weilepp T.G. (1966). Passive length-tension relationship of intact muscle, epimysium, and tendon in normal and denervated gastrocnemius of the rat. Archives of Physical Medicine and Rehabilitation 47, 612-620.

126

Thelen, D.G. (2003). Adjustment of muscle mechanics model parameters to stimulate dynamic contractions in older adults. Journal of Biomechanical Engineering 125(1), 70.

Toussaint, H.M., de Winter, A.F., de Haas, Y., de Looze, M.P., van Dieën, Y.H., & Kingma, I. (1995). Flexion relaxation during lifting: implications for torque production by muscle activity and tissue strain at the lumbo-sacral joint. Journal of Biomechanics 28, 199–210.

Vera-Garcia, F.J., Brown, S.H.M., Gray, J.R., & McGill, S.M. (2006). Effects of different levels of torso coactivation on trunk muscular and kinematic responses to posteriorly applied sudden loads. Clinical Biomechanics 21(5), 443-455.

Wagner, H., Anders, C., Puta, C., Petrovitch, A., Mörl, F., Schilling, N., Witte, H., & Blickhan, R. (2005). Musculoskeletal support of lumbar spine stability. Pathophysiology 12, 257–265.

Walker, S.M., & Schrodt, G.R. (1974). Segment lengths and thin filament periods in skeletal muscle fibres of the rhesus monkey and the human. The Anatomical Record 178, 63-81.

Ward, S.R., Kim, C.W., Eng, C.M., Gottschalk, L.J., Tomiya, A., Garfin, S.R., & Lieber, R.L. (2009). Architectural analysis and intraoperative measurements demonstrate the unique design of the multifidus muscle for lumbar spine stability. The Journal of Bone and Joint Surgery 91, 176–185.

Ward, S.R., Tomiya, A., Regev, G.J., Thacker, B.E., Benzl, R.C., Kim, C.W., & Lieber, R.L. (2009). Passive mechanical properties of the lumbar multifidus muscle support its role as a stabilizer. Journal of Biomechanics 42(10), 1384-1389.

White, A.A., & Panjabi, M.M. (1978). Clinical Biomechanics of the Spine. J.B. Lippincott, Philadelphia, PA.

Wilke, H.J., Wolf, S., Claes, L.E., Arand, M., & Wiesend, A. (1995). Stability increase of the lumbar spine with different muscle groups. a biomechanical in vitro study. Spine 20, 192–198.

Winter, D.A. (2009). Biomechanics and motor control of human movement. 4th ed. Hoboken: John Wiley & Sons Inc.

Winters, T.M., Takahashi, M., Lieber, R.L., & Ward, S.R. (2011). Whole muscle length- tension relationships are accurately modeled as scaled sarcomeres in rabbit hindlimb muscles. Journal of Biomechanics 44(1), 109-115.

127

Woittiez, R.D., Huijing, P.A., Boom, H.B., & Rozendal, R.H. (1984). A three- dimensional muscle model: a quantified relation between form and function of skeletal muscles. Journal of Morphology 182(1), 95-113.

Wolf, S.L., Basmajian, J., Russe, R., & Kutner, M. (1979). Normative data on low back mobility and activity levels. Implications for neuromuscular re-education. American Journal of Physical Medicine 58(5), 217-229.

Zajac, F.E. (1989). Muscle and tendon: properties, models, scaling, and application to biomechanics and motor control. Critical Reviews in Biomedical Engineering 17(4), 359-411.

Zatsiorsky, V.M. (1998). Kinetics of human motion. Champaign IL: Human Kinetics.

Zwambag, D.P., & Brown, S.H.M. (2015). Factors to consider in identifying critical points in lumbar spine flexion relaxation. Journal of Electromyography and Kinesiology 25(6), 914-918.

Zwambag, D.P., & Brown, S.H.M. (2016). Decreasing the required lumbar extensor moment induces earlier onset of flexion relaxation. Journal of Electromyography and Kinesiology 30, 38-45.

Zwambag, D.P., Gsell, K.Y., & Brown, S.H.M. (2016) Variations in slack length and elastic modulus between spinal muscles of the mouse, rat, and rabbit. Canadian Society for Biomechanics Conference Proceedings.

128