Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Cerebrospinal fluid mechanics during and after experimental spinal cord injury 2011

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_2011_fall_jones_claire.pdf [ 33.2MB ]
Metadata
JSON: 1.0071905.json
JSON-LD: 1.0071905+ld.json
RDF/XML (Pretty): 1.0071905.xml
RDF/JSON: 1.0071905+rdf.json
Turtle: 1.0071905+rdf-turtle.txt
N-Triples: 1.0071905+rdf-ntriples.txt
Citation
1.0071905.ris

Full Text

 CEREBROSPINAL FLUID MECHANICS DURING AND AFTER EXPERIMENTAL SPINAL CORD INJURY  by Claire Frances Jones  BSc, The University of Western Australia, 2003 BEng, The University of Western Australia, 2003 MSc, The University of Leeds, 2005   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY  in  The Faculty of Graduate Studies (Mechanical Engineering)   THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) June 2011  © Claire Frances Jones, 2011  ii Abstract Despite concentrated research efforts there is currently no treatment for spinal cord injury (SCI). Several researchers have identified that cerebrospinal fluid (CSF) may have a role in the biomechanics of the injury event and in the secondary physiologic response, but this has not been closely examined.  The aim of this thesis was to develop a large animal model and a benchtop model of human SCI, and to use these to characterise (1) the pressure response of the CSF during the SCI event, (2) the effect of CSF thickness on mechanical indicators of injury severity, and (3) the pressure differentials and cord morphology associated with thecal occlusion and decompression. Study 1 presented the large animal model and provided preliminary CSF pressure transient data that indicated further investigation was warranted.  In Study 2, the CSF pressure transients from medium and high severity human-like SCIs were characterised.  The peak pressures at 30 mm from the impact were within the range associated with experimental traumatic brain injury, but the wave was damped to peak pressures associated with noninjurious everyday fluctuations by 100 mm.  In Study 3, results from the bench-top model demonstrated that the thickness of the CSF layer is directly proportional to the resultant peak CSF pressure, cord compression and impact load. In Study 4, the cranial-caudal CSF pressure differential increased gradually over eight hours of thecal occlusion.  Decompression eliminated or reduced the differential, after which it did not change significantly.  These results indicate that lumbar CSF pressure measured prior to decompression may not be representative of CSF pressure cranial to an injury.  In Study 5, the change in spinal cord and thecal sac morphology after surgical decompression was assessed with ultrasound.  Moderate SCI was associated with a residual cord deformation and then gradual swelling, while high severity SCIs exhibited immediate swelling which occluded the thecal sac within five hours. The different aspects of CSF response to SCI demonstrated in this thesis can potentially be used to assess and validate current and future models of SCI, and to guide future studies of clinical management strategies such as CSF drainage and early decompression.  iii Preface A version of Chapter 2 has been submitted for publication.  Jones CF, Lee JHT, Kwon BK and Cripton PA. Development of a large animal model to measure dynamic cerebrospinal fluid pressure during spinal cord injury.  This study was performed in conjunction with two other studies designed by BK Kwon; I wrote part of the ethics applications.  I was responsible for designing the injury model and pressure measurement methods for the study.  I constructed the injury device, assisted with the animal surgeries, collected and analysed the data, carried out the statistical analysis and was the primary author of the manuscript.  Ethical approval was provided by the University of British Columbia Animal Care Committee under certificates A08-0934 and A08-0935. A version of Chapter 3 has been submitted for publication.  Jones CF, Lee JHT, Burstyn U, Okon E, Kwon BK and Cripton PA. Cerebrospinal fluid pressures during dynamic contusion-type spinal cord injury in a pig model.  I was responsible for designing the study and associated apparatus, and writing the ethics application.  I assisted and/or supervised the surgeries, collected and analysed the data, carried out the statistical analysis and was the primary author of the manuscript.  I received some assistance with histology preparation.  Ethical approval was provided by the University of British Columbia Animal Care Committee under certificate A09-0366. A version of Chapter 4 is being prepared for submission.  Jones CF, Kwon BK and Cripton PA. CSF pressure transients, cord deformation and load transmission are affected by CSF thickness and impact velocity in a bench-top model of contusion type SCI.  I was responsible for designing the study, designing and constructing the experimental apparatus, performing the experiments and collecting the data.  I received assistance with processing the high speed x-ray images.  I analysed the data and was the primary author of the manuscript. A version of Chapter 5 is being prepared for submission.  Jones CF, Newell RS, Lee JHT, Cripton PA and Kwon BK.  The pressure distribution of cerebrospinal fluid responds to residual compression and decompression in an animal model of acute spinal cord injury.  I was responsible for designing the study and associated apparatus, and writing the ethics application.  I assisted and/or supervised the surgeries, collected and analysed the data, and was the primary author of the manuscript.  I received some assistance with statistical analysis.  Ethical approval was provided by the University of British Columbia Animal Care Committee under certificate A09-0366. A version of Chapter 6 is being prepared for submission.  Jones CF, Kwon BK and Cripton PA. Gross morphological changes of the spinal cord immediately after surgical decompression in a large animal model of traumatic spinal cord injury.  I was responsible for designing the study and associated apparatus, and writing the ethics application.  I assisted and/or supervised the surgeries, collected and iv analysed the data, and was the primary author of the manuscript.  Ethical approval was provided by the University of British Columbia Animal Care Committee under certificate A09-0366. v Table of Contents Abstract ........................................................................................................................................................ ii Preface ......................................................................................................................................................... iii Table of Contents ........................................................................................................................................ v List of Tables .............................................................................................................................................. ix List of Figures ............................................................................................................................................. xi List of Abbreviations ................................................................................................................................ xvi List of Symbols ........................................................................................................................................ xvii Acknowledgements ................................................................................................................................. xviii Chapter 1 Introduction ............................................................................................................................ 1 1.1 Overview ............................................................................................................................................. 1 1.2 The spine and spinal cord ................................................................................................................... 2 1.2.1 Anatomical orientations ......................................................................................................... 2 1.2.2 Spinal column ........................................................................................................................ 3 1.2.3 Spinal cord ............................................................................................................................. 5 1.2.4 Functional divisions and cellular components of the spinal cord .......................................... 7 1.2.5 Meninges ................................................................................................................................ 8 1.2.6 Ventricular system ................................................................................................................. 9 1.3 Function, physiology and pressure characteristics of the CSF system ............................................. 10 1.3.1 Functions of CSF ................................................................................................................. 10 1.3.2 CSF formation, circulation and reabsorption ....................................................................... 11 1.3.2.1 Formation locations, mechanisms and rates ........................................................... 11 1.3.2.2 Absorption locations and mechanisms ................................................................... 12 1.3.3 CSF pressure and flow ......................................................................................................... 13 1.3.3.1 Tissue perfusion pressure and pressure-volume compensation .............................. 13 1.3.3.2 CSF pressure ........................................................................................................... 15 1.3.3.3 CSF pulsations ........................................................................................................ 17 1.3.3.4 CSF flow pathways and velocity ............................................................................ 21 1.3.4 Summary .............................................................................................................................. 22 1.4 Human traumatic spinal cord injury ................................................................................................. 22 1.4.1 Epidemiology ....................................................................................................................... 23 1.4.2 SCI classification ................................................................................................................. 23 1.4.3 Spinal fracture and SCI mechanisms ................................................................................... 24 1.4.3.1 Burst fracture .......................................................................................................... 25 1.4.3.2 Dislocation and fracture-dislocation ....................................................................... 27 1.4.3.3 Distraction .............................................................................................................. 27 1.4.4 SCI mechanisms ................................................................................................................... 28 1.4.5 Anatomical risk factors for SCI ........................................................................................... 29 1.4.6 Pathophysiology of SCI ....................................................................................................... 30 1.4.6.1 Vascular changes .................................................................................................... 30 1.4.6.2 Remote diffuse axonal injury in SCI ...................................................................... 31 1.4.7 Clinical treatment options and relevant treatments in research ............................................ 33 1.4.7.1 Perfusion maintenance ............................................................................................ 33 vi 1.4.7.2 Stabilisation and decompression............................................................................. 34 1.4.8 Summary .............................................................................................................................. 37 1.5 Mechanics of traumatic spinal cord injury ........................................................................................ 38 1.5.1 Mechanical properties of the spinal cord ............................................................................. 38 1.5.2 Mechanical properties of the spinal meninges ..................................................................... 41 1.5.3 Rheological properties of the cerebrospinal fluid ................................................................ 43 1.5.4 Mechanics of traumatic tissue injury ................................................................................... 44 1.5.4.1 Mechanical parameters that affect SCI severity ..................................................... 45 1.5.4.2 The mechanical role of CSF in SCI – evidence from animal models ..................... 46 1.5.4.3 Strain and stress tolerance of the spinal cord .......................................................... 49 1.5.4.4 Pressure impulse tolerance of spinal cord .............................................................. 52 1.5.5 Summary .............................................................................................................................. 58 1.6 Modeling human traumatic spinal cord injury .................................................................................. 58 1.6.1 Large animal models for experimental SCI ......................................................................... 59 1.6.2 Relative size of animals for SCI models .............................................................................. 61 1.6.3 Methods of producing injury ................................................................................................ 62 1.6.3.1 Weight-drop ............................................................................................................ 62 1.6.3.2 Controlled displacement and controlled force contusions ...................................... 65 1.6.3.3 Vertebral distraction and fracture-dislocation ........................................................ 66 1.6.3.4 Residual compression models ................................................................................ 67 1.6.3.5 Comparison of the injury strategies ........................................................................ 68 1.6.4 Selection of mechanical input parameters ............................................................................ 71 1.6.5 Synthetic models of SCI ...................................................................................................... 75 1.6.6 Summary .............................................................................................................................. 77 1.7 Measuring CSF pressure in the brain and spine ................................................................................ 78 1.7.1 Quasi-static clinical and experimental pressure measurement ............................................. 78 1.7.2 Dynamic experimental CNS injury pressure measurement ................................................. 79 1.7.3 Summary .............................................................................................................................. 84 1.8 Research objectives and rationale ..................................................................................................... 84 Chapter 2 A Large Animal Model of SCI to Measure CSF Pressure ................................................ 87 2.1 Introduction ....................................................................................................................................... 87 2.2 Methods ............................................................................................................................................ 89 2.2.1 Animals ................................................................................................................................ 89 2.2.2 Injury device ........................................................................................................................ 89 2.2.3 Pressure transducers ............................................................................................................. 92 2.2.4 Surgical protocol .................................................................................................................. 93 2.2.5 Data acquisition, analysis and statistics ............................................................................... 94 2.3 Results ............................................................................................................................................... 95 2.3.1 Contusion injury characteristics ........................................................................................... 98 2.3.2 CSF pressure ........................................................................................................................ 98 2.4 Discussion ....................................................................................................................................... 101 2.4.1 Injury model ....................................................................................................................... 102 2.4.2 CSF pressures ..................................................................................................................... 103 2.5 Conclusion ...................................................................................................................................... 106 Chapter 3 CSF Pressure during Contusion-type SCI ....................................................................... 107 3.1 Introduction ..................................................................................................................................... 107 3.2 Methods .......................................................................................................................................... 108 3.2.1 Animals .............................................................................................................................. 108 3.2.2 Injury device ...................................................................................................................... 109 3.2.3 Pressure transducers ........................................................................................................... 110 3.2.4 Experimental protocol ........................................................................................................ 111 vii 3.2.5 Histology ............................................................................................................................ 112 3.2.6 Data acquisition, analysis and statistics ............................................................................. 112 3.3 Results ............................................................................................................................................. 113 3.3.1 Animals and injury characteristics ..................................................................................... 113 3.3.2 CSF pressure ...................................................................................................................... 114 3.3.3 Histology ............................................................................................................................ 118 3.4 Discussion ....................................................................................................................................... 119 3.5 Conclusion ...................................................................................................................................... 123 Chapter 4 The CSF Layer, Impact Velocity and Mechanical Indicators of Injury Severity ......... 124 4.1 Introduction ..................................................................................................................................... 124 4.2 Methods .......................................................................................................................................... 125 4.2.1 Surrogate cord and dura ..................................................................................................... 125 4.2.2 Physical model and weight-drop device ............................................................................ 126 4.2.3 High speed x-ray ................................................................................................................ 130 4.2.4 Pressure transducers ........................................................................................................... 130 4.2.5 Test protocol ...................................................................................................................... 130 4.2.6 Data and image analysis and statistics ............................................................................... 131 4.3 Results ............................................................................................................................................. 132 4.3.1 Tensile testing of synthetic dura ........................................................................................ 132 4.3.2 Model results ...................................................................................................................... 134 4.4 Discussion ....................................................................................................................................... 142 4.5 Conclusion ...................................................................................................................................... 145 Chapter 5 CSF Pressure Distribution after Acute SCI ..................................................................... 146 5.1 Introduction ..................................................................................................................................... 146 5.2 Methods .......................................................................................................................................... 148 5.2.1 Pressure transducers and drift assessment .......................................................................... 148 5.2.2 Animals .............................................................................................................................. 148 5.2.3 Experimental protocol ........................................................................................................ 149 5.2.4 Data acquisition, processing and statistical analysis .......................................................... 150 5.3 Results ............................................................................................................................................. 153 5.3.1 Pressure transducer drift assessment .................................................................................. 153 5.3.2 CSF pressure and pulse pressure amplitude ....................................................................... 153 5.4 Discussion ....................................................................................................................................... 161 5.5 Conclusion ...................................................................................................................................... 166 Chapter 6 Gross Morphological Response to Decompression .......................................................... 167 6.1 Introduction ..................................................................................................................................... 167 6.2 Methods .......................................................................................................................................... 168 6.2.1 Animals and animal care .................................................................................................... 169 6.2.2 Injury protocol.................................................................................................................... 169 6.2.3 Ultrasound .......................................................................................................................... 170 6.2.4 Image analysis .................................................................................................................... 170 6.3 Results ............................................................................................................................................. 172 6.4 Discussion ....................................................................................................................................... 180 6.5 Conclusion ...................................................................................................................................... 183 Chapter 7 Integrated Discussion ......................................................................................................... 185 7.1 Overview ......................................................................................................................................... 185 7.2 Summary of findings ...................................................................................................................... 185 7.3 Modelling considerations ................................................................................................................ 187 viii 7.4 The role of CSF in the biomechanics of SCI .................................................................................. 191 7.4.1 Peak pressure...................................................................................................................... 192 7.4.2 Pressure impulse ................................................................................................................ 193 7.4.3 Could a CSF pressure transient contribute to SCI? ............................................................ 193 7.4.4 Effect of CSF layer thickness on mechanical descriptors of SCI ...................................... 197 7.4.5 Implications for SCI models in basic science research ...................................................... 199 7.4.6 Implications for clinical research ....................................................................................... 201 7.5 CSF pressure differentials and cord morphology ........................................................................... 202 7.5.1 Discussion .......................................................................................................................... 202 7.5.2 Implications for clinical management of SCI:  CSF drainage and decompression ............ 204 7.5.2.1 CSF drainage ........................................................................................................ 204 7.5.2.2 Decompression ..................................................................................................... 205 7.6 Limitations ...................................................................................................................................... 206 7.7 Recommendations ........................................................................................................................... 209 7.7.1 Improving the injury apparatus .......................................................................................... 209 7.7.2 Defining the mechanical inputs .......................................................................................... 210 7.7.3 Animal model development ............................................................................................... 210 7.7.4 Pressure transients and injury thresholds ........................................................................... 210 7.7.5 Post-injury pressure and spinal cord swelling .................................................................... 211 7.7.6 Improving the bench-top model ......................................................................................... 213 7.8 Contributions .................................................................................................................................. 213 7.9 Conclusion ...................................................................................................................................... 214 References ................................................................................................................................................ 216 Appendix A: Transducer calibrations ................................................................................................... 251 Appendix B: High speed x-ray distortion validation ........................................................................... 252 Appendix C: Pressure transducer drift test .......................................................................................... 256   ix List of Tables Table 1-1  Anatomical orientations and locations ......................................................................................... 3 Table 1-2  CSF formation rates and weight range of various mammals. .................................................... 12 Table 1-3  ASIA Impairment Scale ............................................................................................................. 24 Table 1-4  Summary of relationships between mechanical parameters and observed effect in experimental contusion SCI in various animals. .......................................................................................... 45 Table 1-5  Summary of spinal CSF pressure measurements at time of experimental spinal injury ............ 49 Table 1-6  Summary of fluid percussion experimental TBI which measured intracranial pressure. .......... 53 Table 1-7  Selection of animal models reporting incident pressure and outcome with the fluid percussion injury method. ......................................................................................................................... 54 Table 1-8  Summary of pressures used to induce injury in in vitro neural cell preparations. ..................... 55 Table 1-9  Summary of models reporting incident blast pressure and resultant internal pressure. ............. 56 Table 1-10  Summary of blast injury models reporting incident blast pressure and pathology and/or behavioural outcome. .............................................................................................................. 57 Table 1-11  Summary of pig models of traumatic SCI ............................................................................... 60 Table 1-12  Summary of types of experimental SCI models, the animals used, and the advantages and disadvantages of each. ............................................................................................................ 70 Table 1-13  Canal occlusion velocities for burst fractures at T12 and L1. ................................................. 72 Table 1-14  Comparison of pressure transducers used to characterise dynamic experimental CNS injuries, grouped by type. ..................................................................................................................... 82 Table 2-1  Input injury parameters, numbers of successfully recorded model assessment (output) parameters, and pressure transducer sites by experimental group. ......................................... 95 Table 2-2  Injury parameters and peak positive and negative CSF pressures (relative)  at the “far” location for instrumented SCIs with 50 g and 100 g weight-drop. ....................................................... 97 Table 3-1  Pressure transducer specifications ........................................................................................... 111 Table 3-2  Descriptive statistics for model assessment parameters and results of Mann-Whitney U-tests comparing these parameters for the moderate and high severity injury groups. .................. 114 Table 3-3  Descriptive statistics and results of the Mann-Whitney U-test comparisons for the moderate and high severity injury groups ............................................................................................ 116 x Table 3-4  Results of the Wilcoxon Rank Sum (matched-pairs) tests comparing the cranial and caudal test parameters. ............................................................................................................................ 117 Table 4-1  Elastic modulus (MPa) and thickness (mm) of human, bovine and porcine spinal dura from published data. ...................................................................................................................... 133 Table 4-2  Descriptive statistics for impact velocity, cord and dura diameter, and baseline CSF pressure for each combination of dura size and drop height. .............................................................. 135 Table 4-3  Regression model coefficients for cord compression and impactor, base and tether loads. .... 139 Table 4-4  Regression model coefficients for the peak CSF pressure at each transducer location. .......... 141 Table 5-1  Coefficients and 95% confidence interval for the linear mixed model for blood pressure. ..... 154 Table 5-2  Descriptive statistics for the changes in CSF Pressure at the cranial and caudal location ....... 157 Table 5-3  Coefficients and 95% confidence interval for the linear mixed models for cranial-caudal CSF pressure differential and cranial-caudal pulse pressure amplitude differential. ................... 159 Table 5-4  Descriptive statistics and t-test results for the change in CSF pressure and pulse pressure amplitude at the time of decompression ............................................................................... 160 Table 6-1  Qualitative grading scale for increased parenchymal echogenicity on ultrasound images ...... 172 Table 6-2  Summary of qualitative and quantitative spinal cord morphology and lesion ultrasound grade for each animal. .................................................................................................................... 179  xi List of Figures Figure 1-1  Planes and directions used to describe anatomical positions in the (A) human and (B) quadruped. ................................................................................................................................ 3 Figure 1-2  Sagittal view of the human spinal column (left), superior view of a typical thoracic vertebra (right) ........................................................................................................................................ 5 Figure 1-3  Anatomy of the spinal vertebra and spinal cord. ........................................................................ 7 Figure 1-4  Scanning electron micrograph of the lumbar spinal cord of a 15-month-old child. ................... 9 Figure 1-5  The ventricular system within the human brain. ...................................................................... 10 Figure 1-6  Intracranial pressure volume curve........................................................................................... 15 Figure 1-7  CSF pressure response (central trace) to Valsalva manoeuvre (left) and jugular compression (Queckenstedt test, right), with reference respiration trace (top) and echocardiogram (ECG) trace (bottom). ........................................................................................................................ 16 Figure 1-8  Graph of normal CSF pressure signal with arterial pulsations and respiratory fluctuations (centre), with reference respiration trace (top) and echocardiogram trace (bottom). ............. 18 Figure 1-9  Simultaneous recordings of a single cycle of ECG and CSF pressure pulse in the cerebral ventricle (V), cisterna magna (C), and lumbar subarachnoid space (L), in millimeters of water. ...................................................................................................................................... 19 Figure 1-10  CSF pressure vs. pulse pressure amplitude in 14 healthy volunteers. .................................... 20 Figure 1-11  Lumbar CSF pressure waveforms in a SCI patient before and after decompression. ............ 21 Figure 1-12  SCI etiology (left) and AIS grade at time of discharge (right). .............................................. 24 Figure 1-13  SCI spinal level (left) and spinal column injury (right). ......................................................... 25 Figure 1-14  Medical images of a burst fracture at L2. ............................................................................... 26 Figure 1-15  Distribution of burst fractures in the male and female population by vertebral level. ........... 26 Figure 1-16  Medical images of a bilateral facet dislocation at C6-7 in a 29 year-old male. ...................... 27 Figure 1-17  Lateral radiograph of distraction injury in a 41 year-old male at C5-6. ................................. 28 Figure 1-18  Spinal cord tissue pressure and CSF pressure versus time, after experimental SCI. .............. 37 Figure 1-19  Mean stress-strain curves for uniaxial tensile tests of rat spinal cord. ................................... 39 Figure 1-20  Stress-strain curve of human and bovine lumbar dura. .......................................................... 42 Figure 1-21  Average CSF viscosity and shear stress versus shear rate for one subject. ............................ 44 xii Figure 1-22  Pressure wave in spinal CSF caused by experimental SCI in a cat. ....................................... 47 Figure 1-23  Pressure transients measured in the spinal CSF during a closed column experimental SCI. . 48 Figure 1-24  Functional recovery threshold determinations for two weight-drop SCI models. .................. 50 Figure 1-25  Canal occlusion (mm) and compression force (kN) during an experimental burst fracture in the thoracolumbar spine. ......................................................................................................... 72 Figure 1-26  Percent spinal canal occlusion versus time for a representative in vitro experimental burst fracture (left)[134][134][134][134][134]; Photograph of representative bovine vertebra for estimating anterior-posterior canal diameter (right). .............................................................. 73 Figure 1-27  Injury parameters reported to induce transient or permanent paresis in various large animal weight-drop models. ............................................................................................................... 75 Figure 1-28  Comparing a fluid-filled catheter transducer and indwelling transducer: apparatus schematic and results. .............................................................................................................................. 80 Figure 2-1  Photo of the components of the modified weight-drop injury device. ..................................... 91 Figure 2-2  Schematic of front view (left) and side view (right) of the weight-drop device installed over four vertebrae. ......................................................................................................................... 92 Figure 2-3  Photo showing the surgical site, with 4 pressure transducers implanted intrathecally, a widened laminectomy at the injury site (T10), and pedicle screws in T9 and T12 (top); same photo with overlay indicating locations of pressure transducer tips relative to the injury site (shown as a circle) (bottom). The two near transducers were only implanted for two of the Group C animals. .................................................................................................................... 94 Figure 2-4  CSF pressure: cranial-far (top) and caudal-far (middle), and load (bottom) for the Group B (50 g injury) animals. .................................................................................................................... 99 Figure 2-5  CSF pressure: cranial-far (top) and caudal-far (middle), and load (bottom) for the Group C (100 g injury) animals. .......................................................................................................... 100 Figure 2-6  Plots of CSF pressures for the two tests from Group C in which two “near” transducers were implanted .............................................................................................................................. 101 Figure 3-1  Schematic of front view (left) and side view (right) of the weight-drop injury device installed on vertebrae T10-T13; the injury was centred on the T11 vertebral level. ........................... 110 Figure 3-2  Photo (top) and overlay (bottom) indicating the location of the four intrathecal pressure transducers and pedicle screws; the injury was centred on the T11 vertebral level. ............ 111 xiii Figure 3-3  Typical response for a single injury (#P1805); CSF pressure at four locations, and load versus time. ...................................................................................................................................... 115 Figure 3-4  Peak positive CSF pressure and pressure impulse at each transducer location, for the two injury groups. ........................................................................................................................ 117 Figure 3-5  White and grey matter sparing (%) for the high, moderate and sham animals (left) and cumulative white and grey matter sparing from 8mm cranial and caudal of the epicentre (right). ................................................................................................................................... 118 Figure 3-6  Photographs of microscope sections stained with eriochrome cyanine: (A) High severity injury animal (P1805) at epicentre, 0% tissue sparing; (B) Moderate severity animal (P1697) 3.2 mm caudal of epicentre, 60% tissue sparing; (C) Sham animal (P1628) at epicentre, 100% tissue sparing. ............................................................................................................. 119 Figure 4-1  Schematic of the synthetic spinal cord and dura model (not to scale). ................................... 128 Figure 4-2  Photograph of the experimental setup. ................................................................................... 129 Figure 4-3  Typical high speed x-ray images with the large dura (left), medium (top right) and small (bottom right), immediately prior to contact between the impactor tip and the dura surface.132 Figure 4-4  Stress vs. strain plot for tensile tests on 10 samples (black solid lines) of the plastic used to construct the surrogate dura. ................................................................................................. 133 Figure 4-5  Representative data for one impact (small dura, 32 cm height drop), showing CSF pressure (top), loads (middle) and cord compression (bottom) versus time.  Three frames of high speed x-ray illustrate (A) impactor-dura contact, (B) impactor-cord contact and (C) maximum cord compression. ................................................................................................ 136 Figure 4-6  Cord compression (%) versus drop height for the noCSF condition and the small, medium and large dura sizes. .................................................................................................................... 137 Figure 4-7  Impactor load versus drop height for the noCSF condition and the small, medium and large dura sizes. ............................................................................................................................. 138 Figure 4-8  Base load versus drop height for the noCSF condition and the small, medium and large dura sizes. ..................................................................................................................................... 138 Figure 4-9  Spinal cord tether load versus drop height for the noCSF condition and the small, medium and large dura sizes. .................................................................................................................... 138 Figure 4-10  Peak CSF pressure (mmHg) versus distance from impact epicenter (mm) for each drop height. ................................................................................................................................... 140 xiv Figure 5-1  Example of filtering process for 30 seconds of Cranial pressure data (P1836, t=0-15min). .. 152 Figure 5-2  Linear mixed model and raw data for blood pressure (from cuff measurements) over compression and decompression. ......................................................................................... 154 Figure 5-3  Cranial (red) and Caudal (blue) CSF pressures for 14 hours post-injury, divided into groups exhibiting distinct pre-decompression behaviour (A) consistent and increasing cranial-caudal pressure differential, (B) partial cranial-caudal pressure differential, and (C) little or no cranial-caudal differential. .................................................................................................... 156 Figure 5-4  Linear mixed model and individual animals’ data points for cranial-caudal CSF pressure differentials for periods of (A) Compression and (B) Post-decompression. ........................ 158 Figure 5-5  Comparison CSF pressure (A) and pulse pressure amplitude (B) at the cranial (solid black line) and caudal (dashed grey line) location immediately before (pre-) and after (post-) decompression. ..................................................................................................................... 159 Figure 5-6  Cranial and Caudal CSF pressures for 14 hours post-injury, for the two sham animals (P1611, P1628). .................................................................................................................................. 161 Figure 6-1  Representative post-injury ultrasound image indicating the location of the parameters determined for each image.................................................................................................... 171 Figure 6-2  Pre-injury ultrasound image indicating the location of spinal cord anatomy visible on the ultrasound images. ................................................................................................................ 173 Figure 6-3  Example ultrasound images depicting three typical responses, in three subjects, to decompression following an acute injury with eight hours sustained compression. ............ 175 Figure 6-4  Moderate Injury Severity Animals (panel A-F). .................................................................... 176 Figure 6-5  High injury severity animals (panel A-E) and sham animal (panel F). .................................. 177 Figure 6-6  Ultrasound images showing examples of graded echogenic changes. ................................... 180 Figure 7-1  Injury parameters (height and weight) reported to induce transient or permanent paresis in various large animal weight-drop models, including the parameters used for the studies in Chapter 2 and 3 (red filled markers). .................................................................................... 188 Figure 7-2  Sagittal diameter of human spinal cord (closed markers) and dura (open markers), from vertebral level T1 to T12 ...................................................................................................... 190 Figure 7-3  Axial magnetic resonance image of thoracic spine of 20 kg Yucatan miniature pig (unpublished data). ............................................................................................................... 191 xv Figure 7-4  Peak CSF pressure versus distance from the impact site for the current studies (open markers) and reported values compiled from the literature (filled markers). ...................................... 193 Figure 7-5  Non-injurious (green vertical bars) and injurious CSF/parenchyma pressure transients measured during with experimental SCI and TBI events (vertical bars) compared to median CSF transients measured in pigs at 30 mm and 100 mm cranial and caudal to the injury epicentre for high and medium severity injuries (horizontal lines) (Chapter 3). .................. 195 Figure 7-6  Pressure transients incident on animals or in vitro cell preparations subjected to blast or fluid percussion injury, with evidence of subsequent tissue/cell damage (vertical bars), compared to median CSF transients measured in pigs at 30 mm and 100 mm cranial (upper) and caudal (lower) to the injury epicentre for high (blue) and medium (purple) severity injuries (Chapter 3). .......................................................................................................................................... 196 Figure 7-7  Peak CSF pressure measured at various locations in the current bench-top model and a previous simulation of Hall et al. [462] utilising bovine and surrogate spinal cords and dura (left); Peak CSF pressures measured inside cadaver spinal canals (with [469] and without [473] surrogate cord) subjected to axial impacts (right); compared to median CSF transients measured in pigs at 30 mm (dashed horizontal lines) and 100 mm  (solid horizontal lines) cranial and caudal to the injury epicentre for high (blue) and medium (purple) severity injuries (Chapter 3). .............................................................................................................. 198 Figure 7-8  Bar graph comparing the % cord compression measured during 4.6 m/s impacts for the noCSF, small, medium and large dura cases in the current study (blue bars) (mean±standard deviation), and the % cord compression measured for 4.5 m/s impacts using a 7 g impactor with similar impactor:cord diameter ratio for bovine and surrogate cords with and without CSF (purple bars) (median±standard deviation, where available) ........................................ 199  xvi List of Abbreviations ASIA American Spinal Injury Association BBB Blood brain barrier BSCB  Blood spinal cord barrier CAP Central arterial pressure CCI Controlled cortical impactor CNS Central nervous system CPP Cerebral perfusion pressure CSF Cerebrospinal fluid CSFP Cerebrospinal fluid pressure CSFPPA Cerebrospinal fluid pulse pressure amplitude CT Computed tomography CVP Central venous pressure ECG Echocardiogram EMG Electromyography EP Evoked potential FE Finite element FIR Finite impulse response (filter) FP Fluid percussion HR Heart rate ICP Intracranial pressure IH Infinite Horizon (impactor) IM Intramuscular IV Intravenous LMM Linear mixed model MAP Mean arterial pressure MASCIS Multicenter Animal Spinal Cord Injury Study MR Magnetic resonance NHP Non human primate NYU New York University (impactor) OSU Ohio State University (impactor) PLL Posterior longitudinal ligament PMHS Post mortem human subject RMSSD Root mean squared of standard deviation SCI  Spinal cord injury SCIWORA Spinal cord injury without radiographic abnormality SCIWORET Spinal cord injury without radiographic evidence of trauma SCPP Spinal cord perfusion pressure SD Standard deviation TAAA  Thoracoabdominal aortic aneurysm TBI Traumatic brain injury   xvii List of Symbols F Force [N] g Acceleration due to gravity [m/s2] g-cm The product of grams and centimeters, employed as a “unit” of injury severity in weight-drop SCI models. h Height [m] I Impulse [Ns]  m Mass [kg] ms Milliseconds t Time [s] v Velocity [m/s]  xviii Acknowledgements Many individuals have contributed to the successful completion of this thesis and my broader academic experience at UBC.  I would like to express sincere gratitude to the following people without whom this journey would have been vastly different and not nearly as enjoyable. The mentorship of my supervisory committee has extended far beyond this thesis and provided unique insights into different academic philosophies.  My Ph.D. supervisor, Dr Peter Cripton, provided academic, intellectual and emotional support over many years.  Thank you for the many opportunities you have given me to grow as a researcher and teacher.  I will always treasure your eternal optimism and your friendship.  My unofficial co-supervisor Dr Brian Kwon - from hands-on surgery to sage advice, it always felt like my work was just as important to you as it was to me.  It has been a pleasure and a privilege to work with you and your staff.  Dr Tom Oxland, thank you for your calm and thoughtful direction and support. A small army of people contributed to the animal experiments and tolerated considerable disruption to their lives during that time: the veterinarians and staff at UBC Animal Care, Gordon Gray, Tamara Godbey, Bev Chua, Rhonda Hildebrandt and Kari Jones - thank you for looking after my girls so well and for welcoming me into your workplace; and my animal surgeons, Jae Lee and Uri Burstyn. Past and present members of the Cripton, Oxland, Wilson, Kwon, Hodgson and McKay labs who have contributed, in one way or another, to my academic growth.  Special mention to Robyn Newell, Emily McWalter, Jennifer Douglas, Angela Melnyk, Lindsay Nettlefold, Hannah Gustafson, Laura Greaves, Shannon Kroeker, JD Johnson, Tim Bhatnagar, Carolyn van Toen, Tim Nelson, Chee Leung, Tim Schwab, Qingan Zhu, Chad Larson, Anthea Stammers, Elena Okon, Jessica Hillyer, Melonie Burrows and Danmei Liu. Glenn Jolly and Marcus Fengler, for their timely support and advice at times of electronic and mechanical crisis. Robyn Newell, Peter Cripton, Pete Ostafichuk and Frank Ko, for your various contributions to my teaching, TAing and supervising experience at UBC. Sincere and heartfelt thanks to Robyn Newell and Emily McWalter for their friendship, patience and care, and for providing perspective when it was needed most.  I could not have asked for a better Canadian family - thank you for sharing this journey with me. xix My friends in Perth, and around the world, who have tolerated long silences and always welcome me back as though no time has passed.  Special thanks to Steve, Danielle, Monique, Deb, Esther and Murray. My family, who continue to provide me all things necessary for grounding and success, and the motivation to get back to Australia.  To my parents, Eleanor and Trevor, for believing in me, encouraging me and loving me during this long journey.  My sisters, Christine, Nicole and Michelle, their partners, and little John, Adele and Luke, for their understanding of my absence from their lives.  Special thanks to Christine for her care in Vancouver. Medtronic Inc generously provided implants and surgical instrumentation for the injury device. Project funding was provided by the Canadian Institutes of Health Research and the Natural Sciences and Engineering Research Council of Canada.  The International Collaboration on Repair Discoveries (ICORD), Centre for Hip Health and Musculoskeletal Research, and Vancouver Coastal Health Research Institute provided funding for travel and equipment loans. 1 Chapter 1 Introduction 1.1 Overview The cerebrospinal fluid is an important, but little studied, component of the central nervous system. There is a presumption that it provides mechanical protection for the neural tissue but there is little experimental data to confirm this for the spinal cord.  In contrast, some clinical and experimental evidence suggests that cerebrospinal fluid (CSF) may contribute to the transmission of energy away from the spinal cord injury (SCI) site, leading to remote diffuse injury.  There has also been recent interest in manipulating CSF pressure to control spinal cord perfusion after SCI, but the effects of subarachnoid occlusion caused by SCI and of subsequent surgical decompression on pressure differentials along the spine are not well understood.  This thesis examines the transient response of the CSF pressure during the primary injury event and the trends in CSF pressure and layer thickness in the hours following SCI, using novel large animal and physical bench-top models. The transient CSF pressure response during experimental SCI in animals has been reported by two groups [1-3].  However, these studies reported limited data and used fluid-filled catheters and external transducers which can lead to damping and lengthening of the signal.  Measuring CSF pressure in the spinal region is technically difficult because the fluid layer is very thin, even in humans and large animals.  Spinal catheters generally have a diameter similar to the CSF layer thickness in the cervical and thoracic region and are therefore likely to alter the mechanical response of the tissues and the fluid flow. Some physical models have also measured pressure inside a surrogate cord material during simulations of SCI, but due to the complex geometry and mechanical properties of the central nervous system (CNS) tissues these may not provide a biofidelic response; in addition, none have incorporated the CSF layer. Miniature transducers that permit the sensing element to be implanted in the subarachnoid space have been developed relatively recently.  Combined with a large animal model and a human-like experimental SCI, reasonable measurements of subarachnoid pressure transients at multiple locations along the spinal cord are feasible. SCI often results in residual compression of the spinal cord and occlusion of the CSF pathways (subarachnoid space) due to bony malalignment or fragments and spinal cord swelling.  Early decompression is advocated and CSF drainage has been proposed as treatment options to increase tissue perfusion and reduce secondary ischaemic damage.  Because it is not possible to measure CSF pressure cranial to the injury site in SCI patients, it is not known whether a pressure differential develops over the lesion site in the hours after injury, and if so, if it is then resolved by decompression.  The complexity of the physiological processes that regulate CSF and vascular pressures makes this difficult to predict.  In 2 addition, the time course of spinal cord swelling and hence the patency of the subarachnoid space after decompression has not been studied. The introductory chapter aims to provide an overview of SCI biomechanics, the current state of knowledge regarding the contribution of CSF pressure at the time of SCI, and its behaviour in the hours following SCI with an occluded and decompressed subarachnoid space.  This chapter is presented in seven sections.  First, a basic anatomical review of the spine, spinal cord and associated tissues is given, followed by the functions, physiology and pressure characteristics of the cerebrospinal fluid system. Then the etiology, mechanisms, risk factors and pathophysiology of SCI are outlined, along with two treatment options that are associated with the current work.  Next, a framework for the mechanical aspects of SCI is provided: the properties of the soft tissues and fluid, mechanical parameters influencing SCI severity, CSF pressure transients measured in previous models of SCI and injury thresholds of the neural tissue are outlined.  This is followed by a critical appraisal of past and present methods used to replicate human trauma in experimental SCI models, and the transducers that have previously been used to measure CSF pressure in the CNS.  The chapter concludes with a statement of the thesis objectives and hypotheses which are addressed in the studies presented in Chapters 2 through 6.  Chapter 7 discusses the contribution this thesis makes to understanding of the role of CSF in the injury and hyperacute phase, with recommendations for future research. 1.2 The spine and spinal cord This section provides an introduction to the location and structure of the spinal anatomy associated with SCI, including the spinal column, the spinal cord and its cellular components, the meninges and the cerebrospinal fluid system.  A more detailed introduction to CSF physiology, function and mechanics are provided in the following sections. 1.2.1 Anatomical orientations The three major anatomical planes are the coronal, sagittal, and transverse (or axial) planes (Figure 1-1).  The relative locations of structures are described by five pairs of terms, as defined in Table 1-1. Throughout this thesis, the terms associated with quadrupedal animals are generally used, except the term cranial is substituted for rostral. 3 Table 1-1  Anatomical orientations and locations Human  Quadruped Description Anterior / Posterior Ventral / Dorsal Structures relative to the front/back of the body Cranial / Caudal Superior/Inferior Rostral / Caudal Relative location along vertical axis, towards the head / towards the feet Medial / Lateral Medial / Lateral Proximity to the median sagittal plane / sides of the body Proximal / Distal Proximal / Distal Proximal (closer to a structure’s origin), distal (further from a structure’s origin) Superficial / deep Superficial / deep Relative position with respect to surface of body    Figure 1-1  Planes and directions used to describe anatomical positions in the (A) human and (B) quadruped. Image (A) adapted from Wikimedia Commons.  1.2.2 Spinal column The spinal column houses the spinal cord, giving it structure, suspending it, and protecting it from traumatic loading.  The human spinal column consists of 33 vertebra, which are divided (from cranial to caudal) into the cervical (7), thoracic (12), lumbar (5), sacral (5 fused) and coccygeal (3-4 fused) regions, by virtue of specific anatomical features and function.  The spine has a natural lordotic (anterior convex) curve in the cervical and lumbar regions, and kyphotic (anterior concave) curve in the thoracic region (Figure 1-2). Each vertebra is comprised of the vertebral body, which is the main axial load bearing element, and a series of posterior elements which are attached to the vertebral body via the pedicles.  The posterior 4 vertebral body and posterior elements form the vertebral foramen, which, when stacked together, form the spinal canal in which the spinal cord resides.  Adjacent vertebrae are separated by intervertebral discs, relatively flexible elements which allow intervertebral motion and some compliance for damping loads. The posterior elements comprise the laminae, a number of processes for the attachment of the spinal muscles, and the facet joints which limit intersegmental motion. 5   Figure 1-2  Sagittal view of the human spinal column (left), superior view of a typical thoracic vertebra (right) Graphics adapted from Gray’s Anatomy 1918 (copyright expired).  1.2.3 Spinal cord The spinal cord is categorised into regions and levels corresponding to those of the spinal vertebrae.  Each level controls the functions of a particular region of the body via a defined set of spinal 6 nerves (see below).  The spinal cord is continuous with the brainstem and extends approximately two- thirds of the spinal column from the foramen magnum to approximately the second lumbar vertebra (L2). The inferior end of the spinal cord tapers to the conus medullaris, at the end of which a filament of pial tissue continues to attach to the coccyx.  The spinal cord is roughly oval in cross-section, with the major axis in the coronal plane, and has an increased cross-section at the cervical and lumbosacral enlargements which are situated at the lower cervical and thoracic regions, respectively.  These enlargements correspond to the increased nerve supply to the upper limb (C3–T1) and the lower limb (L1–S3) plexuses (Figure 1-3, right). In axial cross-section, the spinal cord consists of two primary tissue types – the inner gray matter with a distinct butterfly or ‘H’ shape, and the surrounding periphery of white matter (Figure 1-3, left). The grey matter contains neuronal cell bodies which make motor and sensory signals and decisions relating primarily to the functions served by the spinal segment that they reside in, as well as glial cells that support these neuronal functions.  The white matter contains myelinated ascending and descending axons that transmit signals between neurons of the brain/brainstem and those of the spinal cord grey matter, or between spinal levels.  The grey matter is highly vascularised, while the white matter is less so. Communication between the central and peripheral nervous system occurs via 31 pairs of spinal nerves.  Each pair enter and exit the spinal canal laterally at each vertebral level, via the intervertebral foramina (Figure 1-3, left).  Each spinal nerve comprises two bundles of nerve roots, one containing motor fibres that exit the anterior (ventral) cord and the other containing sensory fibres that enter on the posterior (dorsal) cord.  Because the spinal cord is shorter than the vertebral column, the nerve roots become progressively longer and travel further before reaching their intervertebral entry/exit point in the more caudal regions.  The cluster of roots inferior to the tapered end of the spinal cord, are called the cauda equina (Figure 1-3, right). As previously mentioned, each level of the spinal cord controls a particular body region.  Due to the localised functions of the grey matter, damage to a particular level of the spinal cord results in neurological deficits in the body regions associated with that level.  In contrast, damage to the white matter tracts interrupts the passage of signals between the brain and the segments caudal to the level of injury, thus affecting the functions maintained below that level that are controlled by the brain.  7  Figure 1-3  Anatomy of the spinal vertebra and spinal cord. Two vertebrae, each with the inferior intervertebral disc.  The spinal cord, meninges, venous plexus are situated in the spinal canal, between the vertebral body and posterior elements of the vertebrae (left).  The segmental arrangement of the spinal nerves, showing the cervical and lumbar enlargement and the cauda equina in the lumbar region (right). Adapted from Drake et al., Gray’s Anatomy for Students, 2005, with permission from Elsevier:Churchill Livingston [4].  1.2.4 Functional divisions and cellular components of the spinal cord The grey matter can be divided into three functional regions: the dorsal (posterior), ventral (anterior) and intermediate horns; these columns of grey matter extend various lengths along the spinal cord.  The dorsal grey matter contains sensory neurons, while the ventral grey matter contains motor neurons.  The intermediate horn contains sympathetic neurons.  The white matter also has three functional divisions: the dorsal, ventral and lateral columns.  The dorsal column contains axon fibre tracts that carry ascending sensory information from the spinal cord to the brain.  The ventral columns contain axon fibre 8 tracts that carry descending motor information from the brain to the spinal cord.  Subdivisions of the lateral columns have both sensory and motor tracts. In addition to the neurons and axons described above, the glial cells of the spinal cord perform support functions for the neurons and axons.  These include: microglia, which are specialised macrophages that remove damaged neurons and infectious agents; astrocytes, that regulate the extracellular chemical environment; and oligodendrocytes, which form myelin for coating the axons. 1.2.5 Meninges The spinal cord is covered by three membranous layers, collectively referred to as the meninges. The inner most is the pia mater, a thin transparent membrane that is closely adhered to the spinal cord and covers the spinal blood vessels.  The middle layer is the arachnoid mater.  Both the pia and arachnoid membranes are thin, transparent and avascular.  They are connected by extremely thin strands of connective tissue called the arachnoid trabeculae.  The arachnoid is closely adhered to the outermost layer, the dura mater, and these two membranes are commonly referred to collectively as the dura.  The dura mater is a thicker, stronger membrane comprised of an elastic matrix with collagen fibres.  All three membranes extend over the spinal nerves and the brain.  The “space” between the arachnoid and pia, referred to as the subarachnoid space or the intrathecal sac, contains the cerebrospinal fluid, and is continuous with the cerebral ventricles and cerebral subarachnoid space.  It also contains the spinal arteries and veins.  The spinal cord and pia are tethered within the dura/arachnoid by the denticulate ligaments which extend between the interior borders of the subarachnoid space, midway between the dorsal and ventral nerve roots (Figure 1-4).  The spinal dura is located in the spinal canal by attachments to the foramen magnum and to the coccyx via the filum terminale.  The epidural or extradural “space” separates the dura mater from the surface of the spinal canal. It contains variable amounts of epidural fat, the subdural venous plexus, spinal arteries and lymphatic vessels, as well as the ligamentum flavum and posterior longitudinal ligament (PLL) at the posterior and anterior border of the canal, respectively (Figure 1-3, left). 9  Figure 1-4  Scanning electron micrograph of the lumbar spinal cord of a 15-month-old child.  The spinal cord (SC) contains the central canal (CC) and is closely surrounded by the pia mater (P). The dentate ligaments (L) are seen on each side of the cord, and the dorsal septum (S) at the top of the cord; both are continuous with the pia and merge into the arachnoid mater (A). The individual layers of the dura (D) can be seen. The subarachnoid space contains nerve roots (NR) and blood vessels (V). Note that the arachnoid mater is more closely associated with the dura mater and the dural circumference is more oval in shape in vivo.  Adapted from Journal of Neurosurgery, 69(2), Nicholas and Weller, The fine anatomy of the human spinal meninges. A light and scanning electron microscopy study, 276-82, 1988, with permission from the American Association of Neurological Surgeons (AANS) [5].  1.2.6 Ventricular system There are four cerebrospinal fluid filled ventricles in the brain, connected by a series of channels (Figure 1-5).  The two lateral ventricles are within the cerebral hemispheres, and each connect to the more caudal and midline third ventricle via the interventricular foramen of Monro.  The fourth ventricle is located in the midline between the brain stem and cerebellum, and is connected to the third by the cerebral aqueduct of Sylvius.  The ventricles contain choroid plexuses which are capillary rich cellular masses that form CSF, as described in Section 1.3.2.  Cerebrospinal fluid enters the cranial and spinal subarachnoid spaces via several apertures in the roof of the fourth ventricle (foramen of Magendie and two foramina of Luschka), and this ventricle is continuous with the central canal of the spinal cord inferiorly.  10  Figure 1-5  The ventricular system within the human brain.  The locations of the choroid plexuses (solid black structures) and the distribution of CSF (dotted area).  Adapted from Neuroscience, 129(4), Brown et al., Molecular mechanisms of cerebrospinal fluid production, 957-70, Copyright (2004), with permission from Elsevier [6].  1.3 Function, physiology and pressure characteristics of the CSF system CSF is a transparent, colourless, water-based fluid that is formed, circulated and reabsorbed in the ventricles and the subarachnoid space surrounding the brain and spinal cord.  It has a similar composition to blood plasma but contains less protein.  The adult human ventricles and subarachnoid space (cranial and spinal) contain around 25 and 115 mL of CSF respectively, for a collective volume of approximately 140 mL [7].  CSF is formed at a rate of around 0.35 mL per minute [8,9], and thus has a complete volume turnover every 6-8 hours [10].  The functions, physiology and mechanics of CSF are complex but a basic understanding of each is necessary to appreciate how CSF may attenuate or amplify the severity of a SCI, at the time of injury and in the subsequent hours.  This is an area of ongoing research and so this section highlights the current understanding as it pertains to this thesis. 1.3.1 Functions of CSF CSF is commonly ascribed five main physiologic and mechanical functions: (1) providing a chemical environment conducive to the efficient transmission of neural signaling; (2) carrying nutrients and waste to and from the CNS; (3) providing neutral buoyancy for the brain and spinal cord; (4) 11 protecting the neural tissue from contact with the cranium or spinal canal during external body loading; and (5) protecting the CNS from acute blood pressure changes which would alter tissue perfusion, by volume adjustment. This thesis addresses aspects of the latter two items, and an appreciation of the physiology and mechanical characteristics of the CSF pathways and flow is necessary to appreciate the interpretation of the results of the studies herein. 1.3.2 CSF formation, circulation and reabsorption The normal mechanisms, regulators and locations of CSF production and absorption are not well understood.  There is considerably less known about the system’s response to spinal pathology, particularly that resulting from traumatic injury.  A detailed account of current knowledge is given by Johanson et al. [11], and a brief description is given here. 1.3.2.1 Formation locations, mechanisms and rates It is generally agreed that the CSF is formed in the ventricular system in the cranial vault, more specifically originating in the choroid plexus, ependyma and parenchyma.  Quantifying the relative contribution of the sources to total CSF production is technically challenging [12], because considerable operative manipulation is required and unknown compensatory mechanisms may arise [13].  Current agreement is that around 60-90% of formation is choroidal, although estimates have ranged from around 30 to 70% [14,15].  The diffusion of interstitial fluid across the ependyma or pia mater in the brain, may contribute 20-30% of fluid production [13].  Importantly, both of these sources are in the cranial vault. No spinal CSF production has been found in cats [16,17], monkeys [18] or dogs [19]. The choroid plexuses are branched structures made up of villi projecting into the ventricles, where each villus is a single layer of epithelial cells overlying a cluster of connective tissues and blood capillaries.  The CSF is generated by these epithelial cells via a two step process.  The first step is the passive filtration of plasma (ions and water) across the choroidal capillary endothelium, driven by the pressure-gradient between the capillary blood and choroid interstitial fluid.  The second step is active secretion across a single-layered epithelium, which is regulated by a complex array of epithelial transporters and ion channels.  Further details are given by Brown and colleagues [6,20]. The normal CSF formation rate has been determined for several species and appears proportional to species size and therefore probably total CSF volume (Table 1-2).  Formation rate varies with age and this may reflect changes in brain CSF volume [21,22]; it also varies throughout the circadian rhythm [23]. 12 Table 1-2  CSF formation rates and weight range of various mammals.  Weights are as stated by authors unless otherwise indicated.  To the author’s knowledge, there is currently no published data for the pig. Species Weight (kg) CSF formation rate (mL/hr)  [reference] Mouse  0.02-0.04* 0.02±0.004 [24] Rat  0.2-0.4* 0.12 [25] Monkey  2-4 2.1-4.1 1.70±0.15 [18] 1.80-2.65 [26] Cat  2-6* 0.9±0.6 [27] Dog  12-17 2.76±0.12 - 2.82±0.36 [28] Sheep  60-90‡ [29] 25-30 35-40 NR† 2.46±0.42 (young)  1.86±0.44 (middle aged)  1.17±0.16 (old)  [22] 3.37±0.38 [30] 4.94±0.10 [31] 6.67±1.06 [32] Goat  30-60‡ [33]  9.6 [34] Human 25-40* 60-80 60-100* 60-100* 21±1.2  (children) [9] 20.4±7.8 (young adults) [35] 11.4±4.2 (elderly, ~77yrs)  24.6±14.4 (young, ~29yrs) [21] 22.2 [8] *estimated †NR=not reported ‡from alternate paper from same group  Response of the CSF formation rate to CSF pressure is considered negligible under normal physiological conditions.  In short-term experimental elevation of intracranial pressure (ICP) in animals, this effect has been seen in some [36,37] but not all animals [28,34,38] and humans [9,35].  The rate is decreased with significantly elevated intracranial pressure; presumably the initial passive filtration step is diminished due to the decreased choroidal blood flow [13] (see Section 1.3.3.2).  The rate of formation is also thought to be modified by metabolic and other physiological processes, such as osmotic pressure of the blood [12,34] but little is known about these.  The presence of autonomic nerve terminals in the choroid plexus has led some researchers to suggest that there is neurogenic control of CSF secretion [12,39]; however, the functional role of the innervations in normal and pathological conditions is largely unknown. 1.3.2.2 Absorption locations and mechanisms CSF is cleared into the lymph and venous systems at several locations via several physiological mechanisms; recent reviews have been given by Kapoor et al. and Johnston et al. [40,41].  Initially, the cranial and spinal [42] arachnoid granulations and villi were thought to be the sole site for CSF reabsorption into the superior sagittal sinus and epidural veins.  Arachnoid villi are herniations of arachnoid membrane into either the lumen of the superior sagittal sinus in the brain or the small spinal veins adjacent to spinal nerve roots.  More recently, both cranial and spinal lymphatic drainage pathways 13 have been acknowledged [43].  The cranial pathway is primarily through the cribriform plate into the cervical lymph, and the spinal pathway is via lymphatic channels in the dura [41].  Lymphatic absorption accounts for 40-48% of CSF clearance in sheep [30], and there is morphological evidence for similar lymphatic absorption in humans and non-human primates [44]. CSF reabsorption is a passive process driven by a pressure differential between the subarachnoid and venous systems and the rate has a relatively linear relationship to CSF pressure over a wide range of pressures [13,45].  In normal individuals, absorptive capacity is thought to be much greater than that needed to maintain mass balance [38].  Human data have shown that absorption begins at a CSF pressure of 5 mmHg (no venous pressure reported) [9], and in animals reabsorption occurred at a CSF-to-venous pressure differential of 1.5-7 mmHg [46,47].  The pressure differential is crucial to the operation of the arachnoid villi, which act as a one-way valve and allow the passage of CSF into the blood when CSF pressure is higher than venous, but collapse when this gradient is reversed so that blood cannot pass the other way.  Recent evidence suggests that the lymphatic mechanism predominates at lower pressure gradients while the arachnoid villi are secondarily recruited at higher CSF pressures [48]. Spinal absorption has been demonstrated to account for 25% of total absorption in sheep [49] and 50% in cats [27].  In humans, spinal absorption is estimated to occur at around 10 mL/hr [35], which probably accounts for around 50% of total absorption given the formation rates stated above.  Increased spinal absorption is associated with activity compared to resting, in humans [35].  The number of observed arachnoid villi increases from the cervical to the lumbar spine in humans [50], which may imply a caudally increasing absorption rate. 1.3.3 CSF pressure and flow Similar to other physiological pressures, the CSF pressure is pulsatile.  It is described with a mean value and an amplitude value.  The mean value is termed the mean CSF pressure, or simply the CSF pressure, and its characteristics are described in Section 1.3.3.2, below.  The amplitude is most commonly called the pulse pressure amplitude and is described in Section 1.3.3.3, below. 1.3.3.1 Tissue perfusion pressure and pressure-volume compensation The spinal cord perfusion pressure (SCPP) is a measure of the pressure gradient across the capillary bed of the tissue, and relates directly to tissue blood flow and therefore to the supply of oxygen to the tissue’s cells.  It cannot be measured directly and is therefore defined as the difference between the mean arterial pressure (MAP) and the CSF pressure (CSFP), i.e., SCPP=MAP-CSFP.  This relationship shows that, for a given MAP, there is a CSF pressure threshold above which the perfusion pressure is inadequate 14 to prevent tissue ischaemia.  Further, it implies that the perfusion of the tissue can be controlled by manipulating the mean arterial pressure and/or the CSF pressure. The largely incompressible components of the CNS (the brain, spinal cord, CSF and blood) are contained within the rigid bony confines of the cranium and spinal canal, thus an increase in the volume of one component requires an equivalent decrease in the volume of another in order to maintain constant pressure.  This reciprocal compensation is finite and is termed the intracranial compliance.  The phenomenon is most commonly described for the compensation of cytotoxic (extracellular) and/or vasogenic brain swelling.  The cranial venous sinuses are compressed, the CSF is displaced from the cranium to the spinal subarachnoid space and the compliant epidural venous plexus is compressed [51- 53].  The compliance curve, which plots intracranial pressure (ICP against intracranial volume as shown in Figure 1-6, has three phases: (A) adequate compensatory reserves, (B) increasing volume but ICP maintained, and (C) compensatory reserve exhausted.  In the latter phase a small change in volume leads to a significant change in ICP. Although the compensatory systems are described for the cranial space, it is important to note that under non-pathological conditions, the cranial and spinal CSF spaces are communicating, and therefore compensation for increased spinal cord volume could conceivably occur in the opposite direction; i.e. fluid volume may be directed from the spine towards the cranium, or flow into the spinal compartment may be decreased, to maintain constant pressure.  This does not appear to have been discussed in the published literature. 15  Figure 1-6  Intracranial pressure volume curve.  (A-B) good compensatory reserve, fairly constant ICP maintained despite changes in volume; (B-C) poor compensatory reserve, small changes in volume produce large changes in ICP; (C-D) terminal dysfunction of cerebrovascular responses.  Reproduced from Anesthesia and Analgesia, 106(1), Smith et al., Monitoring intracranial pressure in traumatic brain injury, 240-8, 2008, with permission from Wolters Kluwer Health [54].  1.3.3.2 CSF pressure Physiological pressures are usually stated in millimeters of mercury (mmHg), or centimeters of water (cmH2O), where 1 mmHg = 1.34 cmH2O = 0.13 kPa = 0.019 psi.  Clinically, CSF pressure measured by lumbar puncture can form part of the diagnosis of subarachnoid haemorrhage, meningitis, cerebral venous sinus thrombosis, idiopathic intracranial hypertension, and intracranial hypovolaemia [55].  It is measured with the patient horizontal in the lateral decubitus position, and the reference level, or “zero”, is the pressure in the right atrium of the heart.  CSF pressure can also be measured in the ventricles, where it is termed intracranial pressure (ICP); this can also refer to intraparenchymal pressure, but this is not necessarily equivalent to the former [54].  When measured in the supine position, lumbar CSF pressure is the same as ICP in patients with a communicating subarachnoid space [56].  Normal CSF pressure ranges between 5-15 mmHg for the adult [54,57-60], and is lower for children (3-7 mmHg) and infants (1.5-6 mmHg) [54].  The acceptable, non-pathologic, upper limit varies between sources, with the highest being around 18.5 mmHg [58,61]. Considerable transient changes in spinal CSF pressure can be elicited by various everyday activities and clinical tests which temporarily alter the intra-abdominal or venous plexus pressures.  For example, spinal CSF pressure can be altered by nose blowing (18.2 mmHg), breath holding (12.1 mmHg), sniffing (-3.8 mmHg) [62] and coughing (40 mmHg) [63].  The Valsalva manoeuvre, in which the 16 individual exhales forcibly against a closed airway, can elevate CSF pressures to 35 mmHg in normotensive patients [59,64] (Figure 1-7, left).  The Queckenstedt test, which involves acute occlusion of the jugular veins, can induce increases of up to 10.7 mmHg [63] (Figure 1-7, right).  Changes in the diameter of the lumbar intrathecal sac have been visualised radiographically for these two maneouvres [65].  Although these changes can be up to three times the normal pressure, they do not invoke injury during the short period of application.   Figure 1-7  CSF pressure response (central trace) to Valsalva manoeuvre (left) and jugular compression (Queckenstedt test, right), with reference respiration trace (top) and echocardiogram (ECG) trace (bottom). Note that ordinate scales were not provided in the original publication. Adapted with permission from Lakke, Queckenstedt's test; electromanometric examination of CSF pressure on jugular compression and its clinical value, 1969, © Excerpta Medica [66].  Lumbar CSF pressure can differ from ICP when the subarachnoid space is occluded at, or caudal to, the foramen magnum.  This has been observed in patients with a mass lesion or brain swelling that displaces the brain toward the foramen magnum [67], and with transtentorial or tonsillar herniation [68]. Pressure differentials associated with subarachnoid occlusion are the basis of the “spinal block infusion test” [69] and the Queckenstedt test [70], although these appear to have largely been superseded by imaging studies which can visually demonstrate occlusion of the subarachnoid space.  Lumbar puncture in the presence of a complete spinal block due to a spinal tumour can cause “downward spinal coning”, which indicates the presence of a low pressure compartment below the tumour [71].  Further, in a patient with severe cervical spondylosis, a low lumbar pressure measurement returned a false-negative result for hydrocephalus [72], and intracranial hypertension has been noted in several cases of spinal tumours [73- 77].  In cats with a cervical obstruction induced by subarachnoid kaolin injection, CSF pressure was on 17 average 1.5 mmHg higher on the cranial side than the caudal side after four months [78].  Jugular occlusion caused a mean increase in lumbar CSF pressure of 29.7 mmHg in six patients with incomplete spinal block, compared to only 4.2 mmHg in five patients with a complete block [79].  Another implication of obstructed CSF flow is that the volume compensation capacity of each compartment would be restricted to the venous compensation available in that compartment.  Therefore, relatively small changes in tissue volume would likely have a substantial effect on compartment pressure. The pressure-volume relationships of the intracranial and spinal spaces, described in Section 1.3.3.1 above, indicate that CSF pressure is strongly influenced by vascular pressures and volumes, which are in turn influenced by changes in the rates of CSF formation and absorption, osmotic pressure and equilibrium, rates of diffusion and secretion of metabolites, sympathetic nervous system activity and dural elasticity, among other factors.  In short, there are very complex interactions between these fluid systems, which make it difficult to predict the reaction of the system to trauma and the secondary sequelae. 1.3.3.3 CSF pulsations CSF pressure has three predominant pulsatile components: respiratory, vascular and slow wave [54], of which only the first two are applicable to this work (Figure 1-8).  The respiratory pulsation is due to changes in the pressure differential between the subarachnoid space and the pleural or intra-abdominal cavities that occur over the respiratory cycle [80].  These fluctuations occur at a rate of 8-20 cycles/min (0.15-0.35 Hz), with an amplitude of around 0.75-3.75 mmHg for regular breathing and 3.75-7.5 mmHg for deep breathing [81].  This effect is also common to central arterial and venous pressure measurements [82].  The accepted “true” CSF pressure is during the phase of ventilation when the pleural pressure is closest to zero, i.e. at end-expiration [80].  The value may be derived by manual observation of the pressure trace, or by applying digital filtering algorithms to the signal [83]. The vascular pulsation is coincident with the heart beat and therefore has a frequency of 1-1.6 Hz (60-100 beats/min).  Several intertwined mechanisms for transmission of the pulsation from the blood to the CSF have been proposed, and it is likely that a combination of these contribute.  They include: brain and spinal cord motion due to expansion of, and intramural transmission from, the cranial and spinal arteries [84-86], variation in the size of the lateral ventricles [87,88] and choroid plexus expansion [89]. The predominantly cranial location of the aforementioned pulsation sources is consistent with differences observed between the cranial and caudal CSF pulsations measured in both healthy and pathological subjects.  The peak amplitude of the high frequency CSF pulsation is known as the pulse pressure amplitude (Figure 1-8).  In normal patients it is greatest in the ventricles and decreases with caudal distance from the cisterna magna (Figure 1-9) [89,90].  It also has delayed phase in the lumbar 18 region compared to the ventricles and cisterna magna (Figure 1-9) [89].  It is absent in the lumbar region in patients with non-communicating hydrocephalus and spinal subarachnoid blocks [89], and in some SCI patients before decompression [91].   Figure 1-8  Graph of normal CSF pressure signal with arterial pulsations and respiratory fluctuations (centre), with reference respiration trace (top) and echocardiogram trace (bottom).Adapted with permission from Lakke, Queckenstedt's test; electromanometric examination of CSF pressure on jugular compression and its clinical value, 1969, © Excerpta Medica [66].  19  Figure 1-9  Simultaneous recordings of a single cycle of ECG and CSF pressure pulse in the cerebral ventricle (V), cisterna magna (C), and lumbar subarachnoid space (L), in millimeters of water. The CSF waveform exhibits respiratory and vascular pulsations; note the reduced respiratory and arterial pulsations in the more caudal recordings.  The clinical meaning of the pulse pressure amplitude is not fully understood.  Pulse pressure amplitude is influenced by a complex myriad of physiological variables including heart rate, arterial blood pulse amplitude, venous outflow, compliance of the arterial bed and the venous and subarachnoid spaces, and arterial carbon dioxide concentration [92, see comment by Czosnyka].  These variables act in parallel and have interacting regulating mechanisms, so they are difficult to manipulate and test independently.  It is generally accepted that pulse pressure amplitude is proportional to the mean CSF pressure (Figure 1-10), the change in cerebral blood volume over the cardiac cycle and the mean arterial pressure [93-96]. Figure 1.9 has been removed because of copyright restrictions.  The figure was adapted from Figure 3, Archives of Neurology and Psychiatry, 73(2), Bering et al., Choroid plexus and arterial pulsation of cerebrospinal fluid; demonstration of the choroid plexuses as a cerebrospinal fluid pump, 165-72, 1955, American Association of Neurological Surgeons (AANS) [89]. 20  Figure 1-10  CSF pressure vs. pulse pressure amplitude in 14 healthy volunteers.  Data are shown for opening pressures and peak pressures during Queckenstedt’s test (jugular compression). M is for mean, and large bars are standard deviations. Grey dashed line is an estimated line of best fit added by the author.  Adapted from Journal of Neurosurgery, 40(5), Gilland et al., Normal cerebrospinal fluid pressure, 587-93, 1974, with permission from the American Association of Neurological Surgeons (AANS) [58].  It has been suggested that the return of lumbar pulse pressure amplitude may be indicative of successful spinal decompression in traumatic SCI patients [91] (Figure 1-11).  In the cranial space it has been investigated as a prognostic tool for normal pressure hydrocephalus [97] and paediatric hydrocephalus [98], traumatic brain injury (TBI) [99] and subarachnoid haemorrhage [92]. 21  Figure 1-11  Lumbar CSF pressure waveforms in a SCI patient before and after decompression. Note that the respiratory waveform amplitude and pulse pressure amplitude are increased after decompression. Adapted from Journal of Neurosurgery: Spine, 10(3), Kwon et al., Intrathecal pressure monitoring and cerebrospinal fluid drainage in acute spinal cord injury: a prospective randomised trial, 181-93, 2009, with permission from the American Association of Neurological Surgeons (AANS) [91].  CSF pressure in mammals has similar characteristics to humans, but the mean CSF pressure tends to be lower in magnitude.  For dogs in the prone position it was 7.5±1.1 mmHg at C3, and 6.5±0.9 mmHg at L4 [100]; lumbar pressure in the cat was 8.7±0.5 mmHg [101]; primate, 7 mmHg [102]; cranially in the pig, 15±3 mmHg [103]; and in the rat lumbar pressure was 4.2±2.5 mmHg [104].  In the rat cisterna magna, the mean pressure was 4.1 mmHg and rose to approximately 15 mmHg with Valsalva [105]. Several authors have noted that spinal pressure is responsive to changing the angle of the animal’s body [100] and raising the animal’s head [101], both results are a logical consequence of altering the hydrostatic pressure acting on the lumbar measurement location.  Intracranial CSF pressures of up to 100 mmHg in primates [102] and 152 mmHg in pigs [103] have been experimentally induced for short periods without ischaemia or death. 1.3.3.4 CSF flow pathways and velocity Like all fluids within the body, the cerebrospinal fluid is undergoing constant motion to maintain homeostasis.  At any one time, only a small proportion of the total CSF resides in the spinal subarachnoid space; the majority circulates around the brain.  Magnetic resonance (MR) imaging illustrates that the flow of cerebrospinal fluid is pulsatile; at the foramen magnum, during each cardiac cycle there is a short period of cranially directed flow followed by a longer period of caudally directed flow into the spinal subarachnoid space.  While the cranio-cervical junction stroke volume is around 0.8-2 mL per cardiac cycle, the bulk flow volume is estimated to be only 0.032 mL per cardiac cycle [86,88,106]. Various MR protocols and computational fluid dynamics models have been used to estimate CSF flow velocities.  Flow velocity estimated by MR images in healthy adults was 2.0-3.35 cm/s at the cervical spine [107], 0.8-3.0 cm/s at C2 [86] and 0.6-1.1 cm/s at L3/4 [108].  A computational fluid dynamics model estimated peak velocities of 1.25 and 0.19 cm/s at the brain stem and at L1, respectively 22 [106].  The fluid dynamics model of Loth et al. [88] indicated that flow remains laminar throughout the flow cycle (Reynolds number 150-450) and that inertial effects dominate the flow field for physiological flow rates and fluid properties, particularly in the cervical and lumbar regions where the subarachnoid annulus is largest. The velocity with which the normal physiologic CSF pressure wave travels down the spinal canal is generally estimated to lie within the range of 3-5 m/s [106,109,110]; two other estimates have been on the order of 12 m/s but are thought erroneous due to inaccurate material property assignments [111] and errors in assumption of the wave initiation site [110,112]. 1.3.4 Summary CSF formation occurs at various sites in the brain and is regulated by passive and active mechanisms that are not well understood.  There is a bulk flow of CSF from the cranial to the spinal compartment and no CSF formation sites have been detected in the spine of animals that have been tested. CSF absorption back into the lymph and venous systems occurs at both cranial and spinal sites in approximately equal proportions, and is thought to be passively mediated by the pressure gradient between the subarachnoid and venous systems. CSF pressure is determined by many complex interactions that govern the CSF formation and absorption rates, as well as the distribution of pressure and volume within the cranial and spinal vascular systems.  The balance of mean arterial pressure and CSF pressure is critical for adequate perfusion of the neural tissue.  Current knowledge indicates that the CSF pressure cranial to a subarachnoid occlusion could rise to pathological levels if the mechanisms of pressure-volume compensation and formation- absorption regulation are insufficient. CSF pressure has pulsatile components that are respiratory and vascular in origin.  The vascular pulsation, or pulse pressure amplitude, has shown promise as a prognostic indicator for various CNS conditions.  One study has shown that an increase in lumbar pulse pressure amplitude may indicate successful decompression in SCI patients. 1.4 Human traumatic spinal cord injury Although it is a relatively rare occurrence, SCI causes profound and permanent physical disability, most often in young, otherwise healthy individuals.  It has enormous personal, social and economic costs. A thorough understanding of the etiology, primary injury mechanisms and risk factors for SCI is essential to developing successful prevention and treatment strategies.  Studies in animals and humans have generated a substantial body of knowledge regarding the pathophysiology of SCI, however some phenomena remain unexplained.  No pharmacological treatment has yet proven to be efficacious in 23 clinical trials; clinical management strategies such as decompression and perfusion maintenance aim to reduce secondary spinal cord damage, but their effects are not well understood. 1.4.1 Epidemiology The incidence, etiology and trends in SCI are similar across developed countries such as Australia, Canada, European Union, and the USA [reviewed by 113,114].  In general the statistics exclude individuals who die before admission.  Two studies have reported pre-admission death rate of 16% [115,116], while estimates of pre- and post-admission deaths range 15-30% and 4-17%, respectively [117].  In the developed world, the annual incidence of non-fatal traumatic SCI with persistent neurological deficit is estimated at 12.1 to 57.8 per million population [114], which translates to approximately 11,000 new injuries every year in the United States.  Survival after SCI is approximately 95% of the general population for age and gender at 1 year, and 92% at 10 years.  Survivorship is better for the young, females, those with paraplegia and incomplete injuries [118], and those without concomitant TBI [119].  Complications after injury are mostly due to pneumonia, pulmonary emboli, septicemia related to pressure sores and urinary tract infections. SCI overwhelmingly affects males more than females at a ratio of around 4:1 [113,114,120].  Age- specific incidence generally displays a bimodal distribution [114]: approximately half of all SCIs occur between the ages of 16 and 30, and around 10% percent occur at age 60 or older (NSCISC, 2009).  The main causes of SCI are motor vehicle accidents (43%), falls (20%), violence (18%) and sports (10%) [120] (Figure 1-12).  While transport-related SCI remains the highest group, several authors have highlighted an increased incidence of fall-induced injuries, particularly in the elderly population [121,122]. 1.4.2 SCI classification Neurological deficit associated with SCI is classified by (1) the most caudal level with normal sensory and motor function, (2) the completeness of the injury, which refers to the detection of any neurological function caudal to the injury site, particularly in the lower sacral region, (3) the American Spinal Injury Association (ASIA) impairment scale (AIS), and (4) for complete injuries, the zone of partial preservation, which refers to an area between the injury and S5 that retains some motor or sensory function.  The AIS scale is a graded categorisation where AIS A injuries are complete, AIS B through D injuries are incomplete, and AIS E means no neurological deficit (Table 1-3). Nearly half (45%) of all individuals with SCI are classified as AIS A; AIS B and C account for around 10% each and 30% are AIS D (Figure 1-12, right) [120]. 24 Table 1-3  ASIA Impairment Scale AIS Grade Level of Impairment A No motor or sensory function preserved in the lower sacral segments (S4 and S5) B Sensory but no motor function preserved, including the lower sacral segments (S4 and S5) C Motor function present below the injury, and strengths of more than half of the key muscles are graded < 3 of 5 D Motor function present below the injury, and strengths of more than half of the key muscles are graded ≥ 3 of 5 E Motor and sensory functions in key muscles and dermatomes are normal Adapted from the ASIA 2006 Standard Neurological Classification of Spinal Cord Injury Worksheet [123].   Figure 1-12  SCI etiology (left) and AIS grade at time of discharge (right).  Compiled from NSCISC 2009, Table 27 and Table 60 [120].  1.4.3 Spinal fracture and SCI mechanisms The majority of traumatic SCIs occur as a result of dynamic contact between the vertebral column and the spinal cord, usually with failure or loss of integrity of the bone and/or discs.  Insight into the mechanism of the injury, for example the principal loading direction and magnitude, is gained by analysing the injury patterns seen in imaging studies such as radiographs and MR images.  A number of classification systems have been proposed, for example those of Denis [124,125], Magerl [126] and the Spine Trauma Study Group [127,128].  These systems categorise column injuries by the degree of disruption of the anterior and posterior vertebral elements, spinal ligaments and intervertebral discs, and assist the clinician in defining the resultant spinal stability, likelihood of neurological involvement and appropriate treatment paths. 25 Burst fractures and fracture-dislocations each comprise around 30-40% of vertebral fractures, making them the most common type associated with SCI [117] (Figure 1-13).  A smaller proportion of SCIs occur without observable column disruption and are termed SCI without radiographic abnormality (SCIWORA) or obvious radiographic evidence of trauma (SCIWORET).  Other modalities of SCI include laceration injuries which are common to knife and gunshot injuries, chronic myelopathy resulting from canal stenosis or a mass lesion, and ischaemia related to blockage of the vertebral arteries.  The model used in this thesis is thought to mimic most closely the burst fracture and fracture-dislocation injury mechanism which results from an obvious dynamic interaction between the spinal cord and column. Cervical spinal cord injuries are the most common (55%), while thoracic and lumbar injuries account for 35% and 10%, respectively [117,129].  SCIs occur predominantly in the cervical region in older patients, but only 50% have an associated fracture, while SCI in thoracic and lumbar regions are almost exclusively with bony injury [130].  The most common spinal levels of injury are C5 and C6 for tetraplegic individuals, and T12 and L1 for paraplegic individuals [120,131].  Upper cervical region injuries, specifically atlanto-occipital and atlanto-axial dislocations, are often immediately fatal.  Figure 1-13  SCI spinal level (left) and spinal column injury (right).  Compiled from NSCISC 2009, Table 52-55 [120] and Sekhon & Fehlings 2001, Table 5 [117].  1.4.3.1 Burst fracture A burst fracture is characterised by failure of the vertebral body in compression, as a result of axial compression loading such as a head-first impact.  This results in a loss of height of the posterior vertebral body wall, and commonly some degree of retropulsion of bone into the spinal canal, the radiographic 26 hallmark of a burst fracture (Figure 1-14).  Biomechanical studies have shown that the final location of the bone fragment(s) does not reflect the maximum transient canal occlusion and cord compression that occurs during the dynamic event [132-134].  The post-injury sagittal-to-transverse diameter ratio of the canal [135], but not the individual diameters [135,136], of patients with thoracolumbar burst fractures are predictive of neurological impairment.  Burst fractures are most common in the thoracolumbar region [137,138] (Figure 1-15).   Figure 1-14  Medical images of a burst fracture at L2. Axial computed tomography (CT) image showing a lumbar burst fracture with retropulsion of bony fragments into the spinal canal causing spinal cord impingement (A). Sagittal reformatted image in the same patient, showing large amount of bone fragment in the canal (B).  Adapted from European Journal of Radiology, 59(3), Valentini et al., The role of imaging in the choice of correct treatment of unstable thoraco-lumbar fractures, 331-335, 2006, with permission from Elsevier [139].    Figure 1-15  Distribution of burst fractures in the male and female population by vertebral level. Data is from 34 months at a European level-one trauma centre serving a population of 1.4 million.  For clarity, only the vertebrae at the spinal region junctions are labeled.  Adapted from Emergency Radiology, The incidence and distribution of burst fractures, 12(3), 2006, 124-9, Bensch et al., Figure 1, with permission from Springer Science + Business Media [138].  27 1.4.3.2 Dislocation and fracture-dislocation Dislocations and fracture-dislocations involve the anterior-posterior or lateral subluxation of one vertebral body relative to the adjacent one, and are due to a variety of external loading conditions.  They are inherently unstable fractures and commonly include (1) fracture of the facets where the loading is predominantly in the transverse plane, or “jumped” facets where there is a combined flexion and distraction loading that causes failure of the posterior ligaments, (2) fracture of the vertebral arches, lamina and spinous processes due to extension-compression loading, (3) displacement of the superior vertebrae posteriorly and vertebral distraction due to extension-distraction loading.  Dislocation can also occur under torsional and translational loading.  Medical images of a bilateral dislocation with anterior translation of the superior vertebral body and facet joint are shown in Figure 1-16.  Figure 1-16  Medical images of a bilateral facet dislocation at C6-7 in a 29 year-old male. (left) Sagittal CT demonstrates bilateral facet dislocation (white arrow). (right) Three-dimensional reconstruction of CT demonstrates anterior translation of C6 vertebral body relative to C7 (white arrow). Adapted from Western Journal of Emergency Medicine, 10(1), Gomes et al., Bilateral cervical spine facet fracture-dislocation, 19, 2009 [140], Open access journal, no permission required.  1.4.3.3 Distraction Purely distractive traumatic loads, i.e. in the axial direction only, are extremely uncommon.  Cord distraction is thought to occur in concert with dislocation associated with flexion-distraction, extension- distraction and extension-compression loading.  Distraction-type SCIs do not typically occlude the spinal canal and so are thought to occur as a stretching of the cord between points where it is tethered, such as the brain, filum terminalus and nerve roots.  A radiograph of cervical distraction that resulted in a rare complete cord transection is shown in Figure 1-17 [141]. 28   Figure 1-17  Lateral radiograph of distraction injury in a 41 year-old male at C5-6. Demonstrates a 3 cm distraction between the C5 and C6 vertebral bodies (white bracket).  Adapted from The Journal of Emergency Medicine, 25(4), Schauer and Sokolove, Severe cervical spine distraction, 445-7, 2003, with permission from Elsevier [141].  1.4.4 SCI mechanisms Human traumatic SCI mechanisms are generally grouped into three categories: blunt or contusive injuries that result from burst fractures and fracture dislocations, stretch or distractive injuries that are assumed to arise from SCIWORA/SCIWORET, and laceration type injuries.  However, this categorisation currently has limited clinical application or benefit because treatment pathways do not differ according to the status of the spinal cord.  Because the spinal cord has microstructure that is directionally dependent, biomechanics researchers have hypothesised that the primary injury is specific to the direction of the deformation and that this may one day lead to targeted treatments.  Choo et al. have reported that localised patterns of primary injury [142] and secondary pathology [143] were specific to the injury mode in rats subjected to contusion, distraction and dislocation injuries. Bunge et al. classified the pathology of 22 spinal cords obtained from persons who died between 3 hours and 22 years after suffering a SCI [144].  Such pathology probably reflects the energy associated with the injury to some degree, although for the more chronic specimens the late secondary pathophysiological processes might dominate the observed pathology.  In that series, contusion injuries featuring intact surface anatomy but with areas of internal haemorrhage, necrosis and cysts comprised around 50% of cases.  Solid cord injuries, those in which the cord appeared normal externally but diffuse 29 damage was seen on histology cross-sections, comprised 10%, and massive compression injuries (20%) exhibited highly disrupted surface and parenchymal tissue with the lesion epicentre often replaced by scar tissue.  Laceration injuries (20%) were identified by a clean disruption of surface anatomy [144]. 1.4.5 Anatomical risk factors for SCI Individuals with congenitally small spinal canals are thought to be at higher risk of SCI.  The Torg (or Pavlov) ratio was developed as a radiographic measure of spinal stenosis and a tool for predicting an individual’s risk of cervical SCI or neuropraxia, particularly while participating in contact sports.  It is commonly used in the context of play and return-to-play for contact sports with a relatively high risk of cervical SCI.  The Torg ratio is defined as the ratio of the sagittal spinal canal diameter (measured from the middle of the posterior surface of the vertebral body to the nearest point of the corresponding spinal laminar line) to the sagittal diameter of the corresponding vertebral body (measured at its midpoint) [145].  Normalising to vertebral body size is an attempt to account for magnification errors in the radiographs.  It has recently been shown that the measure has high sensitivity but poor predictive value for SCI [146].  This may be because it is based on bony dimensions rather than the dimension of the spinal cord and the spinal meninges.  Advances in CT and MR now enable visualisation of the spinal cord and CSF; Tierney et al. [147] have suggested that an improved metric of SCI risk would quantify the “space available” for the cord as the difference between the mid-sagittal diameters of the canal and spinal cord. The elderly population is generally thought to be at higher risk of spinal fracture and SCI from low energy trauma due to degenerative changes of the spinal column [148].  Spondylotic changes in the geriatric spinal column include disc height collapse, osteophyte growth and hypertrophy of the ligamentum flavum and facet joints [149].  In addition to reduced intersegmental flexibility, spondylotic changes can also lead to canal stenosis and associated chronic myelopathy [149].  The reduction in canal space available for the spinal cord, and potentially a reduced CSF layer around the spinal cord, are thought to predispose the geriatric population to SCI from a relatively low energy trauma, even that which does not cause fracture or ligamentous injury [150]. The notion that congenital or acquired canal stenosis increases the risk of traumatic SCI is logical but difficult to prove.  Further, the emphasis on canal size or canal-to-spinal cord ratio does not consider the role of the CSF layer during the SCI event.  The thickness of the CSF layer varies along the length of the spinal cord, and between different individuals.  Incorporating the CSF thickness relative to the spinal cord and canal size may improve the predictive value of SCI risk metrics.  Because estimating pre-injury tissue dimensions is challenging in SCI patients, biomechanics studies may provide information on the role of the CSF in SCI risk. 30 1.4.6 Pathophysiology of SCI The discussion so far has concentrated on the injury that occurs at the instant of the mechanical insult.  However, SCI is not a single event, but a series of complicated physiologic processes.  The disruption of axons, vessels and cell membranes that occurs during the primary injury initiates a cascade of secondary cellular events which progress and change in a predictable fashion through the immediate (0-2 hours), acute (2 hours - 2 weeks), intermediate (2 weeks - 6 months) and chronic phases (more than 6 months) [151,152].  These secondary processes include vascular dysfunction, edema, ischaemia, cell necrosis, excitotoxicity, electrolyte shifts, free radical production, inflammation and delayed apoptosis [152].  This cascade is thought to be predominantly degenerative, although some responses have been found to be neuroprotective or restorative [152], and the significance of others is unclear.  Experimental pharmacotherapies and other treatments attempt to attenuate, amplify or modify these responses: neuroprotective strategies aim to shield the tissue that escaped injury during the primary trauma but is vulnerable to spreading secondary damage, while neuroregenerative strategies aim to restore tissue that is damaged by the secondary cascade.  Although there has been relatively little research on the primary injury, recent work has shown that the primary cellular injury characteristics vary according to the dominant mode of the spinal cord deformation [143]. The secondary processes are complex and beyond the scope of this work.  This section concentrates on a brief account of local and systemic vascular changes which provide the motivation for the treatments outlined below in Section 1.4.7.  It then discusses some unexplained clinical observations that may be related to the primary injury mechanism that is of interest in this work, i.e. injury mediated by a fluid pressure wave.  These observations are post-traumatic ascending myelopathy and diffuse axonal injury remote to the injury site. 1.4.6.1 Vascular changes Local vascular changes and ischaemia are thought to be among the most important contributors to the secondary injury process [153].  Although the larger vessels such as the spinal arteries are generally spared [152], the primary mechanical insult causes disruption of the microvasculature, which leads to capillary haemorrhage and small vessel thrombosis.  The injury response also generates vasoconstrictors that affect nearby intact blood vessels and promote fluid accumulation (edema) at the injury site.  The combined effect is a profound local hypoperfusion and ischaemia which deprive the neurons and other cells of oxygen and other nutrients.  For contusion injuries, the majority of the vascular changes occur in the grey matter and lead to central necrosis and eventual cyst formation [151]. The local injury response is further compounded by systemic vascular changes.  Spinal cord blood flow is normally regulated by the autonomic nervous system in response to arterial pressure changes, 31 local metabolic requirements, and blood carbon dioxide and oxygen levels [154].  After SCI there is a transient blood pressure increase associated with sympathetic stimulation, but this is quickly followed by a loss of sympathetic nervous system autoregulation, decreased systemic vascular resistance and increased venous capacitance and pooling, with a resultant systemic hypotension and persistent bradycardia [155]. Since local and systemic vascular changes contribute significantly to the secondary biochemical cascade, it is logical that clinical treatments should target haemodynamic support and increase local perfusion.  Currently, haemodynamic support consists of maintaining blood pressure, while local perfusion is increased by decompression surgery or traction.  However, the spinal CSF pressure may interact with these treatments in ways that are not presently understood.  This is discussed in Section 1.4.7. 1.4.6.2 Remote diffuse axonal injury in SCI As previously mentioned, the primary injury mechanisms and some aspects of the secondary injury process of SCI are not well understood.  One area that has not been explored well is the existence of primary injury at some distance from the injury site.  There are several clinically observed phenomena that may provide evidence of primary remote or diffuse damage.  These include cervical SCIs associated with head trauma (without cervical spine injury), remote diffuse axonal injury observed at autopsy, SCIWORA and post-traumatic ascending myelopathy. There have been several clinical observations of SCI associated with TBIs caused by acceleration of the head.  Hadley et al. report cervical spinal cord contusions at autopsy of infants who sustained non- accidental whiplash-type shake injuries, or so-called shaken baby syndrome [156].  Shannon et al. report histological evidence of diffuse axonal injury in the cervical spinal cord of infants who died from shaken baby syndrome [157].  The latter hypothesised that such injury was due to a stretching of the spinal cord due to hyper-extension and flexion of the neck; however, it is possible that such injuries could be contributed to by pressure transients created by the head and brain motion.  Axial strain of the cord has also been suggested to result in non-contiguous SCI [158] and diffuse axonal injury observed with spondylotic myelopathy [159].  The relevance of SCIWORA as an injury classification is currently debated in the clinical community, given the more subtle injuries that can be detected with MR imaging [160].  However,  at least one study reports cases of paediatric SCIWORA with no overt evidence of spinal cord abnormality on MR imaging [161]; a fluid loading mechanism may contribute to such types of injury.  Czeiter et al. showed that impact acceleration TBI in rats evoked traumatic axonal injury in the spinal cord as far away as the thoracolumbar junction [162].  They observed that the majority of the affected axons were close to the surface of the cord and proposed that a shock wave travelling through the CSF at the moment of injury could contribute to an axonal stretch damage mechanism.  The authors 32 report that this was further supported by their unpublished observation of similar damage from fluid percussion induced injuries. In a histological analysis of the spinal cords of 17 patients who died between 30 minutes and 6 weeks after injury, diffuse axonal injury remote from the focal injury site and at up to 24 vertebral levels from the lesion epicentre was observed in all cases [163].  The study does not hypothesise the origin of this damage.  Zwimpfer et al. [164] reported a series of patients with a so-called “spinal cord concussion” in which neurological deficit was associated with a traumatic event resulting in spinal instability, but resolved completely within 72 hours after injury.  The authors likened these transient deficits to brain concussion and postulated that they result from a force transmitted to the cord without direct cord compression. In many individuals, the level and extent of a SCI can improve over the days and weeks after injury, an effect largely attributed to resolution of spinal shock and plasticity of the neural pathways. However, in up to 6% of patients neurological deficit progresses to a higher level, compared to the initial presentation, in the days and weeks after injury [165-167].  Despite being described by Frankel as early as 1969 [168], subacute post-traumatic ascending myelopathy remains a poorly understood condition, likely due to its rarity and difficulty in establishing causation [169].  To be classified as an ascending myelopathy, the neurological deficit must extend at least two levels higher than the initial assessment [169], but can be over entire spinal regions [170].  Changes in MR signal intensity generally occur up to four vertebral levels above the primary lesion [170].  Secondary deterioration has also been noted in patients that presented with cervical spine fractures but were neurologically intact at initial assessment [171].  The condition is associated with increased mortality particularly if the ascension reaches the brainstem [169,172].  Risk factors for delayed or secondary increased neurological deficit related to clinical management have been identified, including: further mechanical insult due to ongoing spinal instability, traction, halo application and Stryker frame rotation, as well as early surgical decompression and failure of haemodynamic support [165-167].  In many cases, however, there is no association with an adverse post-traumatic event, and the exact cause of deterioration cannot be determined [173]. There have been several recent case studies reporting patients with a low level spinal injury having a rapid ascending myelopathy of an unknown origin [170,173-175].  A number of hypotheses for the cause of injury ascension have been proposed, including: decreased perfusion pressure and subsequent ischaemia caused by increased venous pressure [175] or intra-abdominal pressure [174,175]; reperfusion injury after surgical decompression [176]; thrombosis in a major spinal artery leading to arterial hypotension [174,175]; infection [166,169]; and, secondary injury processes such as inflammatory or autoimmune response including apoptosis [170,172,177].  Descending lesions are not commonly reported, probably because they are undetected since they do not increase the level of neurological deficit. 33 Descending myelopathy has been reported for two patients, but was attributed to venous thrombosis and spinal artery occlusion [178]. These delayed ascending, remote, diffuse and non-fracture associated SCIs are not well understood, and it is possible that they are initiated during the primary injury event.  The neurons, axons and glia of the spinal cord could be affected by deformation or over-pressurisation caused by a CSF pressure transient from an event that rapidly deforms the dura but does not impinge on the spinal cord. Biomechanical evidence for this mechanism is discussed in Section 1.5.4.2 below. 1.4.7 Clinical treatment options and relevant treatments in research A considerable number of neuroprotective and neuroregenerative pharmacological strategies have shown promise in pre-clinical studies, but none have proven efficacious in clinical trials.  A detailed discussion of prospective drug and cell therapies is beyond the scope of this work; a number of current reviews are available [152,153,179-181].  However, considerable advances have been made in the critical care, clinical management and rehabilitation of SCI patients and these have led to decreased immediate mortality and morbidity and increased post-injury longevity [118], as well as increased functional recovery and quality of life. Two of the clinical treatment options that have gained support recently are early spinal cord decompression and maintenance of spinal cord perfusion via haemodynamic support.  Both aim to minimise ischaemic damage to the spinal cord by increasing perfusion and oxygen delivery to the affected area and are practiced clinically to various degrees.  Current acute SCI treatment guidelines present these as treatment options rather than treatment standards due to the lack of sufficient scientific evidence to convincingly prove their efficacy [182-184].  The following section discusses the current state of knowledge of each, placing them in the context of this work. 1.4.7.1 Perfusion maintenance Spinal cord perfusion support is currently provided by maintaining adequate arterial oxygenation and blood pressure.  The positive effect of early haemodynamic support, including cardiac inotropes to increase blood pressure, and intravenous fluid volume augmentation/resuscitation, has been recognised in a number of clinical studies [184-186].  Current guidelines recommend maintaining systolic blood pressure >90 mmHg and mean arterial blood pressure between 85-90 mmHg for one week after injury [187].  However, this regime does not consider the contribution of CSF pressure to the perfusion pressure. This is in contrast to clinical management of severe TBI patients, for whom standard care includes monitoring and controlling both ICP and blood pressure [54] to achieve perfusion targets: ICP < 20-25 mmHg, and cerebral perfusion pressure >60-70 mmHg [188]. 34 For TBI patients, ventricular and lumbar drainage has been shown to reduce ICP and increase perfusion, if only transiently [189,190], and patient outcome with prolonged periods of ICP>20 mmHg appears to be impaired [191,192].  CSF drainage is also a recognised strategy for reducing the risk of ischaemic iatrogenic paralysis during thoracoabdominal aortic aneurysm (TAAA) repair in which the spinal cord blood supply is reduced by aortic clamping.  A number of clinical trials and animal studies have demonstrated a reduced risk of neurological deficit or paralysis after TAAA surgery when controlled CSF drainage to maintain a CSFP of 10 mmHg was mandated [reviewed by 193,194,195].  A similar protocol for traumatic SCI has been proposed, and to date has been studied in one animal protocol [196] and one clinical trial [91]. Horn et al. [196] assessed the effect of CSF drainage on spinal cord tissue perfusion, injury severity and functional recovery in a rabbit contusion model.  Following a displacement controlled contusion injury, they drained 0.5-1 mL of CSF from the lumbar subarachnoid space at 1, 2 and 3 hours post-injury to reduce the CSF pressure by 10 mmHg.  While the drainage appeared to reduce the area of tissue damage at the injury site, it did not result in improved electrophysiological or motor outcomes.  Contrary to expectations, they found that spinal cord tissue perfusion (measured by laser Doppler) decreased during intrathecal hypotension. Kwon et al. [91] have recently reported on a prospective randomised trial of lumbar CSF drainage on 22 SCI patients.  CSF was drained to 10 mmHg for 72 hours after surgical decompression, but only when neurological examination was possible and to a maximum of 10 mL per hour.  Lumbar CSF pressure was recorded during the decompression surgery and periodically thereafter.  No adverse events were associated with the drainage, but they did not observe a significant lowering of CSF pressure in the drainage group relative to the non-drainage group. 1.4.7.2 Stabilisation and decompression Stabilisation of the spine and decompression of the spinal cord are recognised as important steps in SCI treatment.  Stabilisation is required to eliminate pathological motion at the injury site and to prevent further neural tissue injury and long-term deformity.  Decompression aims to alleviate residual compression of the cord caused by a malaligned canal or bony fragment.  Decompression can be achieved indirectly by traction, which realigns the spinal canal, or directly by surgical intervention to remove mechanical impingement of the cord.  Despite strong advocacy for early decompression, little work has been done to understand the immediate effects of decompression with regard to the morphological (swelling) response of the cord and the distribution of CSF.  Intraoperative ultrasound is commonly used to assess the adequacy of decompression [197-199], but postsurgical imaging is usually not performed until 24-48 hours later, so it is not known how long the restored epidural and intrathecal spaces remain patent.  A better understanding of spinal cord and CSF response to decompression would likely help to 35 elucidate the benefits of early decompression and perhaps to stratify prognosis by patient response to decompression. While the need for stabilisation and decompression is acknowledged, the role of the timing of surgical intervention is currently uncertain.  To some degree this is because of the difficulty in agreeing on what constitutes “early” and also the considerable disparity between the duration of applied compression in animal studies and reasonably achievable decompression times in the clinical setting. In pre-clinical animal studies, compression duration has ranged from 1 minute to 6 hr; in around half of these studies decompression occurred at 1 hour or less.  In clinical decompression studies the most common definition of early operation is 24 or 72 hours after SCI [200] and only two studies with 8 hr limits are reported [201,202]; reported injury-to-surgery times are mostly longer than eight hours [e.g. 91,203,204], although <8 hr delay times have been reported or are thought to be achievable in some trauma centres [202,205]. Pre-clinical studies using highly diverse animal models have shown mixed benefits in histopathological, electrophysiological, blood flow and behavioural recovery measures.  The following summarises those with more clinically relevant time points; a complete review is given by Furlan et al. [200].  In cats with no spinal cord signal conduction at 6 hrs after a contusion SCI, there was no difference in behavioural outcome for those that were then treated by laminectomy and those that were not decompressed [206].  In rats subjected to a contusion injury and decompressed at 0, 2, 6, 24 or 72 hours, later decompression was associated with worse histopathology, reduced electrophysiological recovery and reduced behavioural recovery at six weeks [207].  Delamarter et al. applied a 60% circumferential compression of the spinal cord at L4/5 in dogs and reported degraded functional recovery and histological findings with compression of 6 hr or longer [208].  Using the same model, Rabinowitz et al. treated dogs with decompression at 6 hr, with or without methylprednisolone, or methylprednisolone while maintaining compression until the two week experimental endpoint.  Surgical decompression at 6 hrs, with or without methylprednisolone, was better than methylprednisolone alone as assessed by electrophysiology, histology and functional neurological improvement [209]. Despite discrepancies in pre-clinical results it is generally accepted that there is “evidence for a biological rationale to support early decompression” [200].  The clinical studies with an 8-hr cutoff showed shorter length of acute care and hospitalisation, less frequent complications, and better neurological outcomes [201,202], but no difference in mortality [201].  Those with later cutoffs have in general provided evidence that early surgical decompression is safe and feasible, can lead to improved neurological recovery and clinical outcome, and reduce the duration of hospitalisation [200]. Current decompression literature lacks an explicit distinction between extradural decompression and subdural decompression.  Edema and haemorrhage can result in a considerable local increase in 36 spinal cord volume, and it is possible that epidural decompression is not sufficient in some cases to reduce parenchymal pressures and improve tissue perfusion.  Increased parenchymal pressure (~20 mmHg) associated with increased water content has been recorded at the injury site two days after a compression SCI in mice [210], and adjacent to the injury site within 30 minutes of injury in cats [211]. Dural decompression (i.e. cutting open the dura) for SCI has been investigated to a limited extent both clinically [212-214], and in animal models [215-217] in the past, but does not appear in SCI treatment algorithms; it is, however, an adjunct procedure in TBI patients with ICP unresponsive to craniectomy [188].  In rats with a mild contusion SCI, durotomy 4 hours after injury was effective in reducing lesion size only when combined with a dural graft [217].  The procedure neither improved or impaired recovery in monkeys with contusion SCI [216].  In humans, the procedure included immediate suturing of the dural incision [214] and debridement of necrotic neural tissue [212,213]; all report positive outcomes but have low patient numbers, limited outcome measures and no control group.  Due to the known variability in spontaneous recovery, these case reports do not permit conclusions to be drawn on the effectiveness of the durotomy procedure. Spinal cord occlusion is due to both bony malalignment and cord edema.  The former is reduced by decompression but little is known about the time course of cord swelling after SCI.  Yeo et al. reported that in three of four sheep receiving a severe T10 injury (50g-20cm), contrast myelography indicated there was no flow past the injury site within one hour of injury, and in two animals this swelling remained for more than 100 hours after injury [218].  Isotope myelography results on one animal were more conservative, with full occlusion at 24 hours reverting to partial occlusion at 44 hours.  Saadoun et al. report complete subarachnoid occlusion due to swelling in forceps compression injured mice at 2 days post injury, as assessed by dye myelography [210]. One animal study has measured pressures cranial and caudal to an experimental SCI, providing information about both the time course of intrathecal occlusion and the effect of SCI on pressure distribution.  Shapiro et al. [211] measured parenchymal and spinal CSF pressures caudal and cranial to the SCI site in cats for at least 4 hours after injury (with no residual compression).  Prior to injury the pressures at each location were equal and the CSF and tissue pressures were not different.  After injury, the pressures cranial to the injury site steadily increased, peaking at around 6 mmHg above baseline 3.5 hours after injury.  The pressures caudal to the injury site decreased steadily until 3.5 hours after injury (Figure 1-18).  The dissociation between the caudal and cranial pressures was interpreted as a loss of communication between the compartments.  In animals that were re-anaesthetised 18 to 24 hrs after injury, the cranial-caudal pressure differentials were resolved and values were similar to baseline, indicating that the subarachnoid occlusion was at least partially resolved in this time frame. 37  Figure 1-18  Spinal cord tissue pressure and CSF pressure versus time, after experimental SCI. Cats were injured by 40g-20cm weight drop. Data points represent the mean pressure for each catheter obtained at 15 minute intervals for 10 animals. Differences between cranial transducers (TP1, TP2) and caudal transducers (TP3, LP) is significant (p<0.001) at 30 minutes after injury. CMP=cisterna magna pressure, LP=lumbar pressure, TP=tissue pressure.  Reproduced from Surgical Neurology, 7(5), Shapiro et al., Tissue pressure gradients in spinal cord injury, 275-9, 1977, with permission from Elsevier [211].  The study of Kwon et al. [91] introduced in Section 1.4.7.1 made an interesting observation regarding CSF pressure before and after surgical decompression in their cohort with 17 cervical and 5 thoracic injuries.  The lumbar CSF pressure increased on average by 7.9 mmHg at the time of decompression, with a mean injury-to-surgery time of 21.6 hrs.  This implies that prior to decompression the CSF pressure cranial to the injury was higher than that caudal to the injury, which in turn indicates that lumbar pressure measurement may not be an accurate measure from which to calculate worst-case perfusion.  At present, cranial CSF pressure monitoring is not indicated in SCI patients without concomitant brain injury and this means that cranial-to-caudal pressure gradients cannot be measured directly in humans.  A better understanding of the spinal cord and CSF response to residual compression and decompression may help to identify a subset of patients which would most benefit from early decompression, or allow more definite prognoses to be made. 1.4.8 Summary Although traumatic SCI is relatively rare, it has devastating and permanent consequences for the individual afflicted.  The primary injury typically occurs from a high energy interaction between spinal canal or bone fragments, and the spinal cord via the dura and CSF.  Burst fractures and fracture- 38 dislocations are the most common injury mechanisms resulting in SCI.  Conditions that increase the stiffness of the spine or reduce the canal-to-spinal cord ratio are thought to increase SCI risk. The primary mechanical injury is rapidly followed by secondary biological processes that further damage the spinal cord, and treatments aim to reduce or eliminate the effects of these mechanisms. Decompression and perfusion maintenance are two clinical treatment options that aim to reduce secondary ischaemic processes.  It is thought that early decompression has a strong biological reasoning and is currently thought to be advantageous in patients who are otherwise stable.  However, the clinical evidence does not equivocally prove efficacy.  Perfusion control is currently based on maintaining a high systemic blood pressure; however, CSF pressure also contributes to the balance of tissue perfusion. Reducing CSF pressure via lumbar drainage has been proposed, but little is known about the distribution of CSF pressure in the cranial and caudal compartments following a SCI.  A better understanding of how the CSF pressure and spinal cord morphology respond to traumatic SCI and subsequent decompression may inform future clinical trials and treatment pathways. 1.5 Mechanics of traumatic spinal cord injury In SCI, the primary damage is caused by the transfer of energy from a moving bone fragment or the canal, through the epidural contents, meningeal membranes and CSF, to the spinal cord.  This causes local tissue deformation and displacement, disrupting neurons and axons, and rupturing blood vessels. Injury or tissue damage is exhibited in a variety of forms, ranging from obvious gross interruption of physical structure to dysfunction or death of individual cells.  The mechanical response of the spinal cord is dictated by the mechanical and rheological properties of spinal cord, dura and CSF.  Therefore, a comprehensive description of these is pivotal to our understanding of the mechanics of SCI and our ability to simulate SCI with computational and physical models. This section also discusses the impact parameters that are known to affect the severity of a SCI and the current knowledge regarding the influence of CSF during the mechanical impact.  Prescribing tissue tolerances is one of the main goals of injury biomechanics; however, it is extremely challenging given the complex mechanical response of biological tissues.  In the remainder of the section, the tolerance of the spinal cord tissue to strain, stress and pressure impulse are described in terms of both mechanical failure and physiologic failure.  This information is largely obtained from the animal models that are described in Section 1.6.3. 1.5.1 Mechanical properties of the spinal cord As described in Section 1.2.3, the spinal cord consists of peripheral white matter surrounding a central core of grey matter.  The spinal cord exhibits the typical soft-tissue nonlinear “J” shaped stress- 39 strain response to uniaxial tension, with stiffness increasing with applied strain (Figure 1-19).  Elastic and tangent moduli have been used to describe the stiffness from 0 to around 10% strain.  In a series of tests using anaesthetised cats and dogs, Hung and colleagues determined an in vivo elastic modulus of around 0.26 MPa up to 5% strain in axial tension with a stretch rate of 0.02 mm/s over one or two spinal levels [219,220].  They also noted increased stiffness due to lack of hydration, lower temperature and increased time after death [221,222].  The effect of perfusion has not been tested in the spinal cord, and studies on the brain have been inconclusive; Gefen and Margulies [223] reported that perfusion did not affect stiffness of in vivo brain tissue, while Weaver et al. [224] found that it affected shear modulus.  Estimates of moduli determined with quasi-static strain rates range from 1.02 to 1.4 MPa for ex vivo human and bovine cord [225-227].  A two-fold increase in elastic modulus at 72 hrs postmortem was reported for ex vivo bovine tissue [227].  Figure 1-19  Mean stress-strain curves for uniaxial tensile tests of rat spinal cord.  Specimens were preconditioned at the strain indicated in the legend. The curves exhibit a toe-region followed by stiffening at higher strains. Note the increased stiffness at the higher strain rates (filled symbols).  Reproduced from Journal of Biomechanics, 38(7), Fiford et al., The mechanical properties of rat spinal cord in vitro, 1509- 15, 2005, with permission from Elsevier [228].  Similar to other soft tissues, the hyperelastic stress-strain behaviour in axial tension is thought to originate from the crimped or undulating fibrous microstructure of the longitudinally oriented axons [229] and the collagen and elastin of the associated vascular tissue and connective membranes.  In addition, this preferential alignment of axons is thought to contribute to anisotropic material properties [225,226], but this has not been verified experimentally [230].  The elastic moduli of white and grey matter were independent of direction when assessed by the pipette aspiration method [231]; however, this method tests a highly localised section of tissue and therefore has limited contribution from non-cellular components.  The elastic moduli stated above were for spinal cords tested with intact pia mater (the pia is closely associated with the cord and is difficult to remove), and this may contribute to directional material 40 properties.  The pia mater increased the stiffness of human cervical spinal cords tested less than 48 hours after death by fifteen-fold [225].  Tests using tissue cores have recorded tangent moduli for white and grey matter ranging from 0.03 to 0.112 MPa [232,233], which is substantially lower than for intact cords with pia mater.  Since the neurons and cell bodies within the grey matter are randomly oriented, the grey matter may be less stiff than white matter in axial tension.  When tested in axial tension, core specimens of grey matter had higher moduli (0.1 MPa at 15% strain) than white matter (0.038 MPa at 15% strain) [233], but this difference was not detected in a study that used the pipette aspiration method on a small section of tissue [231].  At the cellular level, spinal cords with demyelinated axons and/or loss of glial cells were less stiff and had lower tensile strength than controls, indicating that myelin and the glial matrix contribute to the tensile response [234]. Transverse tensile loading and axial and transverse compression loading have not been extensively investigated.  The elastic modulus for rabbit spinal cord in transverse tension was estimated at 0.016 MPa with the pia, and 0.005 MPa without pia.  Ozawa reported greater stiffness in cords with pia, compared to those without, when transverse compression exceeded 1 mm or around 30% of the anterior-posterior diameter [235].  Porcine white matter exhibits nonlinear [236] and rate dependent stress-strain behaviour in axial compression [237].  Considerably more work has been done on the compression and shear properties of the brain, which has a similar cellular makeup but lacks anisotropy [reviewed by 230].  The majority of SCI animal models apply either quasi-static or dynamic transverse compression loads, but in general they are not appropriate for determining material properties since the geometry of the tissue undergoing compression cannot be well defined.  For example, Hung et al. [238] demonstrated the variation in “overall moduli” from zero to 40% transverse compression for a single in vivo cat spinal cord without dura.  However, the modulus was defined as the ratio of stress to compressive strain, as per the tensile definition, which does not directly translate to transverse compression of a cylinder.  The combination of inverse finite element model analysis and materials testing for prescription of material properties of very soft biological tissues has been advocated [239], and a recent study has implemented such a scheme to derive constitutive models of spinal cord white matter [240].  The inverse finite element approach may prove helpful for deriving mechanical properties from transverse compression tests in the future. Due to its high water content, the spinal cord displays highly viscoelastic properties.  The loading and unloading profiles in uniaxial tension exhibit hysteresis, indicating strain history dependence of the stress-strain response [219-222,238,241].  Ex vivo uniaxial tensile tests have shown that stiffness increases with strain rate for intact pia-covered human [226], adult and neonatal rat [242,243] spinal cords, and biopsy specimens of bovine white and grey matter [232].  One study did not detect a difference with strain rate for human specimens with and without pia mater [225].  Several studies have reported stress relaxation properties; higher initial strain and strain rate produced higher stresses at the end of 41 relaxation [226,232,242,243].  The relaxation period tracked in these experimental protocols has varied markedly from 25 sec [226] to 30 min [228].  Several groups have derived constitutive equation constants for various viscoelastic material models [226,242,243]. Few studies have tested intact spinal cords to failure to obtain ultimate tensile stress and strain. Biopsy specimens of bovine white matter failed at 0.061 MPa and 126% strain, compared to grey matter at 0.043 MPa and 50% strain [233].  The lower ultimate tensile stress and strain of grey matter compared to white matter may be consistent with the pattern of central grey matter damage and peripheral white matter sparing that is observed in human and experimental SCI [152].  Rat spinal cords (with pia) failed at 12% strain and 0.08 MPa [228], and chick embryo cords failed at 42% strain and a stress of 0.085 MPa [234].  In vivo cat and dog cords attained a (quasi-static) strain of 40-50% before the grips slipped, without tissue failure [219-221,241]. The variability in results due to viscoelasticity, postmortem degradation and sensitivity to hydration, temperature and perfusion are compounded by difficulty in defining cross-sectional area and in gripping the specimen.  Most tensile tests use the intact cord rather than cutting test coupons and therefore approximate the cross-sectional area as that of an ellipse with major and minor axes corresponding to the lateral and anteroposterior diameters.  It is difficult to provide end conditions that grip the sample without slippage and that do not affect the tissue behavior and introduce end effects.  Differences in preloading and preconditioning protocols also contribute to variation within and between studies [244].  There may also be natural variation by species and age. 1.5.2 Mechanical properties of the spinal meninges Like other connective tissues, the spinal meninges are a composite of collagen and elastin fibres in extracellular matrix material and have non-linear elastic properties.  The elastic and viscoelastic tensile mechanical properties have been studied for human dura [245-253], and that of several mammals including bovine [249,254,255], non-human primate [256], canine [257-259] and rat [260].  In all cases the dura was indistinct from the arachnoid mater for testing and the majority of studies have concentrated on lumbar dura only.  Only one study has tested the spinal pia mater [235]. There is considerable variability in mechanical properties among species and studies.  This can be partly attributed to differences in experimental protocol, such as strain rate, preconditioning, environmental conditions, storage conditions, and end fixation.  In general the dura is one or two orders of magnitude stiffer than the spinal cord, with elastic modulus ranging from around 1 to 150 MPa across the species listed above.  The failure stress ranges approximately 1.4 to 28.5 MPa, and failure strain is between 33 and 100%.  The elastic modulus of rabbit pia mater, calculated from the difference in elastic 42 modulus obtained in tensile tests of spinal cord with and without pia, was 2.3 MPa, approximately 460 times spinal cord parenchyma with the same test protocol [235]. The spinal dura is frequently assumed to be transversely isotropic [257] due to longitudinal orientation of the collagen fibres which should, according to the remodeling patterns of other connective tissues, result from being predominantly loaded in the axial direction during flexion, extension and lateral bending of the spine.  Longitudinal collagen alignment has been found in some microstructural studies [248,249,260] but has not been detected in others [261,262].  One study has reported a tendency for preferential longitudinal and transverse alignment of elastin, rather than collagen, fibres [261].  The ultimate tensile stress and elastic modulus were higher in the longitudinal than the transverse direction for human [248,249,253] (Figure 1-20) and porcine dura [263], but the ultimate strain did not depend on loading direction [248,249] (Figure 1-20).  Figure 1-20  Stress-strain curve of human and bovine lumbar dura.  Strain was applied in the longitudinal direction for human (H/L) and bovine (B/L) dura, and in the circumferential direction for human dura (H/C). The grey shaded region was used to calculate the elastic modulus (E). Region A-B denotes the toe-region in which little stress is applied to obtain a large strain.  Reproduced from Anesthesia and Analgesia, 88(6), Runza et al., Lumbar dura mater biomechanics: experimental characterisation and scanning electron microscopy observations, 1317-21, 1999, with permission from Wolters Kluwer Health [249].  Studies on the dependence of tensile mechanical properties on dorsoventral side and spinal region have had mixed results.  No difference in failure stress and strain, and elastic modulus, was found for rat dura [260], the same was found for human except that elastic modulus increased slightly with caudal progression [250].  Reduced failure stress and strain with caudal progression was observed for dorsal, but not ventral, human dura [246].  For porcine dura, transition and failure strains decreased towards the 43 lumbar region, while elastic modulus and failure stress increased with caudal progression [263].  Dorsal failure stress was lower than ventral, and the opposite relation applied for failure strain, for all spinal regions for human and porcine dura [246,263].  Elastic modulus was lower for ventral specimens than for dorsal specimens of porcine dura [263]. 1.5.3 Rheological properties of the cerebrospinal fluid CSF is a transparent, colourless fluid, derived from blood plasma.  It is mostly water, and in the nonpathological individual contains small amounts of lipids, electrolytes, enzymes, vitamins, amines, sugar proteins, and blood cells.  The composition may be altered by disease or injury that affects the blood-brain or blood-spinal cord barrier (BBB, BSCB). CSF is a Newtonian fluid with a constant viscosity of 0.71 – 0.76 mPa.s at 37 °C [264].  Protein content has a slight but insignificant effect on viscosity [264,265].  Bloomfield [265] compared distilled water with CSF obtained from the ventricular shunts of 23 adult patients with hydrocephalus, pseudotumour cerebri and subarachnoid haemorrhage.  Using identical protocols and on the same rheometer, the absolute viscosity ranged 0.66-0.98 mPa.s and 0.65-1.09 mPa.s, for the water and CSF samples respectively (Figure 1-21), and there was no dependence on protein, glucose or red blood cell concentration.  The average specific gravity (i.e. density relative to water) of CSF from the same patients was 1.007 (range 1.0062 – 1.0082) [265].  Both Bloomfield et al. [265] and Brydon et al. [264] used CSF obtained from patients undergoing procedures for CNS pathology.  Bloomfield et al. confirmed that there were higher than normal levels of protein, glucose and red blood cells in some patients, and normal levels of albumin, immunoglobulin-G and white blood cells in all patients [265]; Brydon et al. only reported protein content [264].  To the author’s knowledge there is no study of CSF rheology using samples from a healthy normal population.  The above properties imply that for the purpose of a mechanical model which does not attempt to replicate biochemical environment, distilled water is suitable to simulate CSF. 44  Figure 1-21  Average CSF viscosity and shear stress versus shear rate for one subject.  Data points represent the mean and standard deviation of 3 replicate tests at each shear rate, for one CSF sample from a single subject.  The rectangles represent the range of viscosities obtained for distilled water (dashed line, filled in green) and CSF (solid line, filled in pink) in the same study (n=23), using the same viscometer for shear rate 360-460 s-1. Tests were done at 37 °C.  Adapted from Paediatric Neurosurgery, 28(5), Bloomfield et al., Effects of proteins, blood cells and glucose on the viscosity of cerebrospinal fluid, 246-51, 1998, with permission from S.Karger AG, Basel [265].  1.5.4 Mechanics of traumatic tissue injury The mechanical response of biological tissue is normally tested under controlled conditions which restrict the applied loading to one mode (tension, compression or shear) and the loading vector to one of the anatomical planes.  This ideal loading is quite different from the loads applied to the spinal cord during a SCI.  The models introduced in Section 1.6.3 have been used to determine the parameters of the mechanical insult that affect the extent of tissue damage.  For the most part these studies assess injury severity by qualitative or quantitative assessment of neurophysiology, patterns and extent of cellular damage as seen on histology, immunohistochemistry and medical imaging, and the functional recovery of surviving animals according to animal-specific scales that rate the performance of various motor skills [266].  Such assessments are generally not of the primary injury, probably because in the immediate post- injury phase histopathology is less developed, MR imaging abnormalities can be absent and functional assessments are dominated by spinal shock [152].  However, the severity and evolution of the secondary processes are likely proportional to the severity of the primary insult and damage. 45 1.5.4.1 Mechanical parameters that affect SCI severity It is well established that the severity of a SCI is dependent on the magnitude of the mechanical parameters which describe the impact.  These parameters include peak force and displacement, impact and deformation velocity, impact energy and load impulse.  For example, slow compression or stretching of short duration may not injure the cord, but prolonged or rapid deformations do [267].  Since the extent of the primary injury dictates the severity of secondary processes, the relative importance of different mechanical parameters has been tested. This has been done predominantly in the mode of rapid transverse compression.  For these contusion-type impacts, the force-displacement profiles are dictated by the mechanical impact parameters as well as the material and structural properties of the system.  It is therefore impossible to evaluate the absolute independent contribution of each to the injury outcome. Table 1-4 highlights some of the findings relating mechanical parameter to injury outcome; these have been determined using the animal models discussed below in Section 1.6.3. Table 1-4  Summary of relationships between mechanical parameters and observed effect in experimental contusion SCI in various animals. Mechanical parameter Observed effect Peak force • Inversely proportional to locomotor recovery scores and spared tissue (but also associated with increased displacement) [268] Peak displacement (compression) • At high peak forces, displacement is proportional to myelin loss, macrophage response and ventral motor neuron loss and inversely proportional to motor recovery [268] • Inversely proportional to amplitude of evoked potentials [269,270] • Inversely proportional to white matter sparing and motor recovery [271] • At low velocities, tissue damage is more dependent on degree of compression than rate of application [272] • Correlates best with later locomotor scores in mice [273] Velocity • Impact velocity and rate of compression are proportional to haemorrhagic necrosis [269] • Proportional to behavioural recovery and tissue damage [267] • Proportional to blood-spinal cord barrier disruption [274] • In rats, contusion at 300 mm/s gave haemorrhage extending into the peripheral white matter, but at 3 mm/s vascular damage was limited to grey matter [275] • Smaller weights falling from greater heights (higher impact velocity) were associated with less haemorrhage, edema, axonal disruption and myelin fragmentation, compared to larger weights from lesser heights [276] Impact energy  • NOT proportional to injury severity (see Section 1.6.3.1) Impulse (∫Fdt) • Proportional to lesion volume [277] • Proportional to intramedullary haemorrhage [278] • Correlates best with early (day 1) locomotor scores in mice [273] VC (velocity x  %compression) • At higher impact velocities, damage to neural tissue is proportional to the time-varying viscous response [272] • VCmax =0.91, corresponds to 50% probability of full recovery [279] • VCmax =1.41, corresponds to 50% probability of partial recovery [279]  46 1.5.4.2 The mechanical role of CSF in SCI – evidence from animal models The ability of the CSF to protect the spinal cord from contact with the spinal canal during large external loads is commonly stated in general reference texts.  However, quantitative and qualitative analysis or theory of the mechanisms by which this occurs are somewhat lacking in the literature.  Hung et al. [1] showed that when CSF was drained from the subarachnoid space, although the overall deformation of the dura-cord-system was similar to that with the CSF, its deformation relative to the initial diameter was greater.  A similar response was shown with ex vivo and synthetic models of SCI [280,281].  However, none of these studies were able to visualise the spinal cord deformation through the natural or synthetic dura.  On observing that altering the hydrostatic pressure of the CSF by draining some or all of it did not change the force-time or energy-time trajectories of the impact, Hung et al. further hypothesised that there was some threshold for the impact parameters above which the “shock-absorbing mechanism of the dural sac and spinal fluid” was ineffective [1].  TBI studies have shown that removing modest amounts of CSF increases brain rotation relative to the skull [282] and decreases the head impact energy required to produce cerebral concussion [283].  Ommaya suggested that the CSF might reduce shear stresses in the brain and spinal cord by damping their movement [284].  However, the brain and spinal cord differ considerably in their geometry, mass, tethering and physical response to external loading mechanisms, so observations about the brain may not be directly transferable to the spinal cord. In contrast to the protective role, several authors have hypothesised that CSF may contribute to focal and diffuse injury when the external forces are larger than those experienced in everyday life.  They propose that a pressure wave travelling away from the point of mechanical insult may transmit injurious loads to the spinal cord [1,3,162,285].  In Section 1.5.4.4 below, it is shown that the neural tissue can be damaged by fluid pressure impulses acting directly on the dura or transmitted through the skull.  Further, as described in Section 1.4.6.2, the causes of post-traumatic ascending myelopathy and diffuse axonal injury remote to the primary lesion site are not known.  In addition, although SCIWORA and SCIWORET injuries are commonly attributed to hyper-flexion, -extension or -distraction of the spine, this has not been confirmed.  It is possible that transmission of the CSF pressure wave during the SCI event may contribute to these primary and secondary injury phenomena. Only two groups have measured spinal CSF pressure during an experimental SCI (experimental SCI models are discussed in Section 1.6.3, below).  Hung and colleagues [1] reported a peak positive pressure of 50 mmHg approximately 2.5 cm cephalad to the injury site in a single cat using the classic open weight-drop model with a 15 g, 5 mm diameter cylindrical weight dropped from 25 cm.  The pressure trace provided indicates that the positive pressure pulse was followed by a smaller negative pressure pulse to approximately -16 mmHg (Figure 1-22).  In a separate publication, using a similar 47 weight-drop cat model, they reported a 150 mmHg peak pressure at 1.5 cm caudal and cranial to the injury site, in response to a 20g-15cm injury [2].  Figure 1-22  Pressure wave in spinal CSF caused by experimental SCI in a cat.  Adapted from Surgical Neurology, 4(2), Hung et al., Biomechanical responses to open experimental spinal cord injury, 271-6, 1975, with permission from Elsevier and The World Federation of Neurogurgical Societies [1].  Using the closed weight-drop dog model described below in Section 1.6.3.3, Wennerstrand et al. [3] report a positive and negative pressure peak of up to 1960 mmHg and 630 mmHg, respectively, at 6 cm from the injury site for a single animal (see curve 3, Figure 1-23, left).  They also noted that the peak positive and negative pressures decreased with distance from the impact site (Figure 1-23, right).  At 15 cm distance these pressures had mean values of 399±89.7 mmHg and 288.2±224 mmHg (N=5), respectively, and at 44 cm, 27.5±6.6 mmHg and 19.0±6.5 mmHg (N=4).  In a fluid percussion brain injury model (see Section 1.5.4.4) using rabbits, a similarly rapid damping of fluid pressure was observed [286].  The peak amplitude and duration of the pressure pulse decreased with distance from the impact site, and while cranial pressures of up to 3000 mmHg were measured, no pressure changes occurred caudal to the fifth cervical vertebra. 48  Figure 1-23  Pressure transients measured in the spinal CSF during a closed column experimental SCI.   Pressure wave patterns recorded in the spinal canal on transverse impact at the L2-L3 region. Distance between the impact site and gauge points: Curve 1: 36 cm cranial; Curve 2: 8 cm cranial; Curve 3: 6 cm cranial (left).  Relationship between the maximum CSF pressure and the distance between impact site and the position of the pressure transducer (right).  Reproduced from Journal of Biomechanics, 11(6- 7), Wennerstrand et al., Mechanical and histological effects of transverse impact on the canine spinal cord, 315-31, 1978, with permission from Elsevier [3].  Two studies have measured spinal pressures during experimental whiplash simulations. In pigs subjected to whiplash-type neck flexion-extensions, pressures recorded in the cervical spinal canal ranged from -100 to 150 mmHg [287,288].  The authors observed that higher pressures were associated with the location of nerve roots in which the neuron cell body membrane was damaged.  The same group placed catheter-tip pressure sensors subdurally in the cervical spine of post-mortem human subjects undergoing whiplash sled tests and recorded pressure amplitudes between 0 and 220 mmHg [289].  Since the vascular and CSF systems were not pressurised in these tests it is unclear how these pressures relate to the in vivo situation. All of these studies observed both positive peaks and negative troughs in the pressure measured at a single location.  Wennerstrand et al. hypothesised that the negative pressure indicated a tensile loading in the spinal cord and dura [3].  Krave et al. observed similar characteristics in pressure traces in the brain and speculated that negative pressures causing tensile loading on the tissue are more likely to cause injury through a stretch mechanism [290].  A study on the sensitivity of brain tissue to the application of suction during neurosurgery found that negative pressures of 75-110 mmHg were sufficient to cause blood brain 49 barrier dysfunction and tissue injury [291].  Negative pressures applied with various vacuum techniques have been shown to induce neural tissue damage and are described in Section 1.5.4.4 below. Several physical models have also been used to estimate the pressure in the spinal cord, but not the CSF, at the time of simulated SCIs.  The construction of these models and pressure results obtained from them are discussed in Section 1.6.5. In summary, there is limited data available on the CSF pressure transients produced during SCI. The data from experimental contusion SCIs was obtained over four decades ago and utilised transducers that were external rather than situated in the subarachnoid space (Table 1-5). Table 1-5  Summary of spinal CSF pressure measurements at time of experimental spinal injury  Injury model type Animal Measurement location Pressure (mmHg) Hung 1975 [1] Open contusion (weight-drop 15 g, 25 cm) Cat Spinal CSF (2.5 cm caudal of epicentre) -16 to 50 Albin 1975 [2] Open contusion (weight-drop 20 g, 15 cm) Cat Spinal CSF (1.5 cm cranial/caudal of epicentre) 150 Wennerstrand 1978 [3] Closed fracture-dislocation Dog Spinal CSF ~0 cm from impact 10-15 cm 15-45 cm  > 2000 1000 – 400* 400 – 30* Bostrom 1996 [287] Svensson 1993 [288] Whiplash (rapid head flexion/extension) Pig Spinal canal (cervical) -100 to 150 Eichberger 2000 [289] Whiplash (rapid head flexion/extension) PMHS† Spinal CSF (cervical) 0 – 220 *Order indicates decreasing peak pressure with distance from the impact site. †Post mortem human subject. Note that the CSF and vascular systems were not pressurised.  1.5.4.3 Strain and stress tolerance of the spinal cord In addition to the gross mechanical tissue failure thresholds discussed in Section 1.5.1, “failure” of neural tissue can be defined by permanent functional and behavioural deficit, permanent or transient reduced signal conduction, mechanical failure at the cellular level and cellular dysfunction.  Since these mechanisms are common to all neural tissue, many experiments have been carried out with relation to TBI; however, the review below is predominantly limited to spinal cord specific experiments. Functional recovery, the rate at which the experimental animal regains motor and sensory ability and the level which is achieved prior to the experimental endpoint, is assessed with a variety of animal- specific instruments such as the Tarlov [292] and Basso-Beattie-Bresnahan [293] scales.  Despite the extensive use of contusion type injury models (predominantly weight-drop, see Section 1.6.3.1) and 50 functional recovery scales in large animals, surprisingly few studies have done extensive exploration to determine the thresholds for permanent injury.  Macaque monkeys recovered full function within 72 hours with a 20g-10cm weight-drop injury but had irreversible paraplegia with a 20g-15cm injury [294]; in a common marmoset model with a drop height of 5 cm, full recovery occurred with a 15 g weight, residual upper limb paralysis occurred with 17 g, and almost complete loss of motor function occurred with 20 g [295].  Ford et al. present one of the most comprehensive examinations of the injury threshold for functional recovery in their cat model, and found that 20g-10cm was the threshold for zero recovery at 6 weeks post-injury (Figure 1-24) [296].  A similar analysis by Wrathall et al. showed permanent loss of hind limb weight bearing and locomotion in rats with the OSU weight-drop device (see 1.6.3.1) with a 10g-17.5 cm injury at T8 (Figure 1-24) [297].  Figure 1-24  Functional recovery threshold determinations for two weight-drop SCI models. (left) Plot and tabulation of final recovery score versus injury height for cats, all with a 10 g weight. Control animals are shown at 0 cm height.  Reproduced from Journal of Neurosurgery, 59(2), Ford, A reproducible spinal cord injury model in the cat, 268-75, 1983, with permission from the American Association of Neurological Surgeons (AANS) [296].  (right) Frequency distribution of final (4 week) motor score for each injury height using the OSU weight-drop device for rats, N=9 or 10 for each height. Reproduced from Experimental Neurology, 88(1), Wrathall et al., Spinal cord contusion in the rat: production of graded, reproducible, injury groups, 108-22, 1985, with permission from Elsevier [297].  Although only a small number of human SCIs occur in “pure” axial tension via the vertebral distraction mechanism, tensile loading is most likely a component of all SCI mechanisms.  The effect of tensile loading on functional recovery derives from the tensile tests of Hung and colleagues that were used to determine in vivo mechanical properties discussed in Section 1.5.1.  In dogs subjected to quasi- static strain of 10-50% strain over L1-L2, those subjected to 10% could stand within 12 hrs and had full recovery at 3 days, while 50% strain had full motor and sensory recovery at 5 days [221,241].  Similar tests on cats showed improvement within one day for strain <20%, within two days for 30-50% strain, and full recovery within 5 days for all of these animals [219].  It is not known how these strain 51 magnitudes compare to the strains experienced in traumatic SCI;  further, the quasi-static loading rate is likely to influence the tissue damage and functional recovery in tension, as it has been shown to do in transverse compression (see Section 1.5.4.1). Failure of the in vivo spinal cord to conduct action potentials occurs prior to gross mechanical failure and is due to either axon deformation or lack of tissue perfusion.  Action potential conduction is assessed by measuring the amplitude of evoked potentials (EP).  Nacimiento et al. applied rapid transverse compression to cat spinal cords and noted a marked decline in EP (to 74% of the uncompressed EP level) at 60% compression; compression to 80% further reduced EP to 36% and compression to 100% abolished EP completely and irreversibly [270].  During quasi-static distraction of the cervical spine in monkeys to a column strain of 8.2% and estimated cord strain of 3.7±1.7%, no evoked potential fell below 37% of control value [298]; 95% reduction is considered necessary for unrecoverable SCI in the cat and monkey [299,300].  The strain threshold for electrophysiologic impairment of the guinea pig optic nerve (strain rate 30-60/s) was 18% [301].  The membrane potential of single squid giant axons was disrupted transiently by low-rate strains up to 19%, increased strain rates caused greater and longer lasting membrane potential changes, and axons elongated past 20% strain did not recover resting potentials [302].  Action potential conduction was abolished in excised strips of guinea pig spinal cord white matter at around 100% strain, and signal recovery was dependent on degree of initial stretch [303].  In quasi-static stretching of the spinal column of dogs, spinal cord blood flow dropped to 27% of control and evoked potentials were eliminated, at an interstitial cord pressure around 46.5 mmHg (the corresponding strain was not reported) [304]. Isolated cell or cell bundle preparations have been used to determine the mechanical response of axons independent of connective tissue and vasculature.  Excised unmyelinated squid giant axons fail structurally at around 25% strain when stretched at a rate of around 10/s [302].  A higher strain tolerance was reached in a cell culture preparation of axons bridged between two rows of neurons; there was no severing of axons observed at <65% strain with a strain rate of 25-35/s [305].  Axonal stretch tolerance of the guinea pig optic nerve, which is essentially a bundle of long parallel axons, at 30-60/s, was similar to the squid axon; the reported strain threshold for morphologic damage was 21% [301]. Although beyond the scope of this review, it is important to note that mechanical insult can also lead to non-catastrophic damage to the plasma membrane of cells.  This “mechanoporation” does not cause cell death but transiently or permanently affects cell integrity and function, disturbing ion homeostasis, electrical activity and signalling [306].  It is thought that this may have implications for secondary damage processes [306].  Experiments in this area typically apply stretch, pressure and shear loads to in vitro cell preparations; several reviews are available [306-308]. 52 The discussion above has largely been limited to measures of strain; stress tolerance in compression is more difficult to ascertain because the internal stress fields are likely to be different from the stress at the interface between the tissue and the compressing platen.  The stress tolerance of spinal cord tissue has been estimated using an approach that combines finite element and animal models. Ouyang et al. [309] measured electrophysiology during quasi-static compression of ex vivo guinea pig spinal cords and determined the patterns of cellular damage with histology.  These patterns were then correlated with the tissue stress derived from a finite element model subjected to the same mechanical deformations, and it was estimated that acute mechanical axon damage occurred at a von Mises stress of around 2 kPa. 1.5.4.4 Pressure impulse tolerance of spinal cord As described in Section 1.5.4.2, pressure transients have been measured in the spinal CSF of animals during experimental SCI.  These pressure transients may damage the spinal cord via gross deformation of the tissue or via microscopic deformation of the cellular membranes.  To the author’s knowledge, there is no established pressure tolerance value for the bulk spinal cord or its cellular constituents.  However, several techniques that are used to study TBI induce the injury using fluid pressures that are transmitted through the CSF, without direct contact with the brain.  These studies provide evidence of the fluid pressure magnitudes and durations that are sufficient to elicit an injury response in neural tissue.  It is recognised, however, that tissue tolerance can be specific to the conditions of the external loading, geometry of the anatomy and local tissue differences such as vasculature, metabolic processes and cellular distribution.  The three TBI models described in this section are the in vivo (animal) and in vitro (cell culture) fluid percussion methods and the in vivo (animal) blast injury method. The fluid-percussion model is one of the most frequently used TBI models [310,311].  It produces brain damage or dysfunction by applying a transient pressure impulse to the intact dura via a fluid-filled chamber attached to the animal’s skull.  Rapid acceleration of a piston at the opposite end of the chamber produces an impulse in the fluid, which introduces a small volume of fluid into the epidural space and produces local elastic deformation of the brain [312].  The impulse is applied either centrally or lateral to the midline via a craniotomy.  The resultant pathology is a mixture of focal and diffuse injury [311]. Percussive impacts are graded in terms of the peak applied pressure, and so this may provide evidence for pressure thresholds of neural tissue.  However, the incident pressure is usually measured immediately upstream of the cranial attachment, and because the dura is stiffer than the neural tissue, extradural pressures are likely to be higher than the subdural pressures incident on the brain.  As summarised in Table 1-6, three studies have shown that intracranial pressure transients measured adjacent to the injury are similar but smaller than those measured extracranially; only one of these placed the intracranial 53 transducers subdurally [313-315].  At measurement sites remote from the injury, peak pressure and impulse are attenuated, but still proportional to, the incident pressure [313].  Table 1-6  Summary of fluid percussion experimental TBI which measured intracranial pressure. Studies are presented in chronological order. Pressures have been converted to mmHg regardless of units used in the study. All pressures are relative. Author Animal Impulse location Extradural (incident) pressure (mmHg) Transducer location Peak pressure (mmHg) Sullivan 1976 [315] Cat Central 304-3800 Extradural, Supratentorial (adjacent to impulse) 304-3800 Stalhammar 1987 [313] Cat Central 990† 2200† 3800† Extradural, Supratentorial  (adjacent to impulse) / Infratentorial‡ 990 / 800† 2280 / 2130† 4180 / 3800† Clausen 2005 [314] Rat Lateral 1440 ± 76 1275 ± 76 Ipsilateral ventricle Contralateral ventricle 1328 ± 228 1380 ± 228 * NR=no reported, but publication from same group found that incident pressure similar to extradural pressure [286] † Selected data, estimated from plots ‡ Transducers placed against unopened dura § NR=not reported. Authors report device can deliver 0-25 bar but do not report the incident pressure for the selected data shown.   Despite the difficulties in defining the actual pressure incident on the brain, the incident pressures that have been applied give an estimate of pressures required to produce neurological injuries of different severities.  Some examples in different animals are shown in Table 1-7 below.  The pressures resulting in a moderate to severe TBI typically range from 1000 to 3000 mmHg, with an impulse duration of around 20 ms.  Mild injuries are reported for pressures as low as 300 mmHg [312,316,317].  Considerably higher pressures were tolerated in sheep than the other animals [318], but the reason for this is not clear.  54 Table 1-7  Selection of animal models reporting incident pressure and outcome with the fluid percussion injury method. Studies are presented in order of ascending animal size. Author Animal Impulse location Pressure (mmHg) Outcome/severity rating    Intracranial (parenchymal) Walter 1999 [319] Swine Lateral 3040-3420§ (anterior) 2050-2200§ (middle) (10-15 ms duration) NR*    Extracranial (incident) McIntosh 1989 [316] Rat Central 75-760 1140-1520 2280-2660 w/o ventilation 2280-2660 w/ ventilation (21-23 ms duration) Slight neurological deficit at 4 weeks Moderate neurological deficit at 4 weeks Severe neurological deficit at 4 weeks Severe neurological deficit at 4 weeks but less than w/o ventilation Delahunty 1995 [320] Rat Central 1520 ± 38 Moderate; transient neurological suppression and persistent motor and memory deficit Rinder 1969 [317] Rabbit Central 140-690|| 310-960 600-1850 no concussion slight concussion severe concussion Hartl 1997 [321] Rabbit Lateral 2625 (20-25 ms duration) Increased ICP, white blood cell activation Sullivan 1976 [315] Cat Central 1140  1672 1976  2432 3040 No EEG‡ changes; physiological response but no pathology Physiological response but no pathology EEG‡ recovery severely impaired. Macroscopic lesions Extensive basal subarachnoid haemorrhage Haemorrhagic contusion at impact site Hayes 1987 [312] Cat Central 0-680† 760-2960† 3040-3800† Mild (microscopic SAH‡ and petechial IPH‡) Moderate Severe, irreversible Marmarou 1990 [322] Cat Central Lateral 2165 2165 4 of 7 survived at 24 hr 5 of 11 survived at 24 hr Hilton 1993 [323] Cat Central 1350 ± 152 2075 ± 53 No midbrain or brainstem haemorrhage Lesion area 33.8±7.3 mm2 Millen 1985 [318] Dog Central 2400 2225 2235 <6 min physiologic change <25 min profound electrical activity depression Increased plasma catecholamine concentration Pfenninger 1989 [324] Piglet Central mean 2700, peak 12000 (18 ms duration) Increased ICP, decreased CPP and cerebral blood flow up to 2 hr Armstead 1994 [325] Piglet Lateral 1397-1691 (12-23 ms duration) Decreased cerebral blood flow and oxygenation Fritz 2005 [326] Piglet Lateral 2354 ± 441 3016 ± 736 Immediate increased ICP (both groups) Focal pathological damage Diffuse axonal injury; second ICP peak @ 5 hr post-injury w/ reduced cerebral blood flow Millen 1985 [318] Sheep Central 3085, 4940,5700,6460,7600 No physiological change * NR=not reported, w/=with, w/o=without 55 † Reported values were pressure impulse [atm.ms]. Approximate pressure values were derived from impulse assuming a triangular impulse with duration of 20 msec ‡ EEG=electroencephalogram (brain electrical activity), SAH = subarachnoid haemorrhage, IPH=intraparenchymal haemorrhage. § Estimated from plots, transducers were used outside of their manufacturer calibrated range || Publication from same group found that incident extracranial pressure was similar to intracranial (extradural) pressure [286]  As noted in Section 1.5.4.2, negative pressures have been observed during experimental SCI in animals.  Several studies have applied negative pressures to the intact dura or directly on the pia to induce focal brain lesions but not diffuse injury associated with significant long-term deficits [310].  Fluid vacuum pulses of 1520-6080 mmHg and less than 100 ms duration applied to the exposed cortical surface of rats caused focal hemorrhagic lesions without overt damage to other regions [327,328].  There was blood-brain barrier breakdown at 10 min [328], and neuronal cell loss, hypertrophy of astrocytes and axonal damage immediately adjacent to the lesion at 3 days [327].  Vacuum pressures of -700 mmHg applied dynamically to the intact dura of rats and held for 5 seconds caused subarachnoid haemorrhage, neuronal damage, astrocyte response, BBB breakdown and changes in cerebral blood flow of various levels at 5 minutes to 7 days [329,330]. Transient fluid pressure impulses have been used to produce injury in in vitro cell cultures mounted on a rigid substrate, as summarised in Table 1-8; the techniques are reviewed by [307].  Cell cultures of neuronal and glial cells exposed to fluid pressure impulses of 1550 mmHg [331] and 362-3050 mmHg (duration 20-30 ms) [332], respectively, show cell damage and reduced viability.  Single or double pressure impulses of 3800 mmHg were used to induce damage in cultured rat neurons, astrocytes and microglia [333,334]. Table 1-8  Summary of pressures used to induce injury in in vitro neural cell preparations.  Studies are presented in chronological order. Author Cell culture Incident pressure /duration  (mmHg / ms) Pathology Suneson 1989 [331] Neuronal (rat) 1550 Cell damage, reduced viability Shepard 1991* [332] Glial (human) 362 / 20-30 3050 / 20-30 Cell damage, reduced viability Wallis 1995  Hippocampal slices (rat) Weight-drop 1kg-61cm  Severe neuronal dysfunction Panickar 2002* [334] Neurons, Astrocytes, microglia (rat) 3800 Free radical production Jayakumar 2008* [333] Astrocytes (rat) 3800 / 25  twice Cell swelling, mediating chemicals identified * Same apparatus used.  Non-impact blast wave induced neurotrauma can occur from exposure to a high pressure air wave resulting from an explosion.  The exact mechanism of energy transmission to the brain is unknown, but 56 one pathway is through the skull, dura and CSF [335].  The pathological response to blast includes many features of SCI, including diffuse edema, metabolic disturbance and vasospasm, neuronal swelling, axonal pathology and white matter degeneration.  Models of blast injury expose the animal to a blast wave produced by detonation of an explosive charge, or release of compressed air in a shock tube [336]. Two studies have measured the pressure at various locations in a rat [337] and swine [338] brain during an experimental blast event (Table 1-9).  In these animals the peak pressure reached inside the brain tissue was similar to the exposure pressure outside the skull.  While it should be recognised that energy transmission probably differs by species head size and skull thickness [335], this implies that the incident pressures are a reasonable estimate of the peak pressures occurring in the CSF and brain and that are causing injury. Table 1-9  Summary of models reporting incident blast pressure and resultant internal pressure. Author Animal Incident pressure / duration (mmHg/ms) Transducer location Resultant peak pressure (mmHg) Pathology/ Physiology/ Behaviour Chavko 2007 [337] rat 315 / 4.5 3rd ventricle 300 / 4.5 -50 / NR* No injury Bauman 2009 [338] swine NR* Ipsilateral: Fore- and hind-brain, thalamus; Contralateral: thalamus NR**. Approximately half the peak magnitude of the incident extracranial pressure White matter degeneration, astrocytosis; vasospasm * NR=not reported. **Due to USA Department of Defense restrictions pressure values were not reported and graphs were provided without scales in the paper [339].  Moochhala et al. [340] found that rats exposed to 21 mmHg blast pressures showed no behavioural or histological response, but those exposed to 150 mmHg (2.5 msec duration) had a significant decrement in physical tests and degenerating neurons in the cerebral cortex of the brain.  Rats exposed to low level blast overpressures of 75-450 mmHg had impaired cognitive function [341] and pigs exposed to 338 mmHg pressure had small haemorrhages and signs of edema in the brain [342].  However, in contrast no injury was detected in the visual neural pathways of rats exposed to a 625 mmHg blast [343].  Exposure to pressure levels of 750-2250 mmHg caused injury to neuronal and glial cells, as well as brain edema in rats [343-349], and 1500-2250 mmHg caused transient apnoea in pigs [350].  Kato et al. [351] applied blast overpressures of 7500 mmHg to  rats with craniotomies.  Despite the considerably larger incident pressure than the aforementioned studies, they observed no significant haemorrhage in cortical or subcortical regions, but did note mild morphological changes such as spindle-shaped changes of neurons and elongation of nuclei toward the shock wave source at 24-hrs post shock.  They suggested that the threshold for blast induced brain injury may be lower than for other organs due to an “intrinsic vulnerability of neurons” and the delicate neural vasculature.  From this synopsis it is clear that the 57 severity of the injury incurred is dependent on the configuration of the experimental blast apparatus and the location of the animal within it.  However, these values do provide an indication of the magnitudes of pressures that might be expected to cause damage to the neural tissues.  A summary is provided in Table 1-10. Table 1-10  Summary of blast injury models reporting incident blast pressure and pathology and/or behavioural outcome. Author Animal Incident pressure (mmHg) / duration (ms) Pathology, physiology and/or behavioural outcome Kaur 1995 [344] rat NR / NR Increased microglial activity Petras 1997 [343] rat 625 / NR 780-825 / NR 970-1300/ NR No injury No injury (2) / injury (2) Injury Axelsson 2000 [350] pig 1500-2250 / 2.5-3.5 Transient apnoea and flattening of EEG (n=4) No significant change in cardiac physiology Saljo 2000 [345] rat 1155 / NR 1800 / NR changes in the neuronal cytoskeleton Saljo 2001, 2002a,b [346-348] rat 225 / NR 450 / NR Activate microglial cells and astrocytes; elevated immunoreactivity; neuronal apoptosis; change ICP and cognitive function Saljo 2003 [352] rat 139 / 2.5 197 / 2 Diffuse neuronal and glial cell damage Moochhala 2004 [340] ‡ rat 21 / NR 150 / 2.5  No injury Neuron degeneration and physical test decrement Kato 2007 [351] rat* 7500 / NR 75000 / NR Mild cell morphology changes Contusional haemorrhage, neuronal apoptosis Saljo 2008 [342] pig 340 / NR Small brain haemorrhages; signs of brain edema Saljo 2009 [341] rat 75-450 / NR Impaired cognition Svetlov 2010 [349] rat 825 / 2 1275 / 4 2685 / 1 1290 / 4 2685 / 10 Transient agitation† Lethal† Lethal† NR Focal massive haemorrhage; diffuse and focal mild neuron damage; BBB disruption * With craniotomy † Body not armoured ‡ Specifically concentrated on central visual pathway and terminal targets in midbrain and diencephalon EEG=electroencephalogram, NR=not reported 58  1.5.5 Summary The anisotropic and viscoelastic mechanical properties of central nervous system tissues, as well as the complex geometric structures of the individual components, make accurate prediction and modeling of mechanical response to external loading difficult.  Traumatic injury to the spinal cord occurs when physical deformation resulting from direct contact with misaligned or fractured parts of the spinal canal exceeds the structural or functional tolerance of the tissue.  The mechanical impact parameters influencing SCI severity include the impact velocity, the peak displacement and load, the rate of deformation, impulse and the velocity-compression product. Since SCI arises from a transmission of kinetic energy to the spinal cord via the CSF, fluid pressure may contribute to cellular and vascular damage in SCI.  Two groups have measured CSF pressure transients during experimental SCI, but it is unclear how well the SCI replicated human injury.  A limited number of animals were tested and fluid-filled catheters, which potentially damp the pressure signal, were used.  A fluid pressure tolerance for the spinal cord and brain likely exists but the tolerance limit is as yet unknown (or unpublished). In vivo and in vitro fluid percussion and blast injury models of TBI have shown that fluid impulses can interfere with CNS function.  The incident pressures used to create mild and moderate brain injuries are within the range of those that have been measured in the spinal CSF during experimental SCI.  Therefore, investigation of CSF pressure transients generated during human- like experimental SCI and measured with indwelling pressure transducers is warranted. 1.6 Modeling human traumatic spinal cord injury Spinal cord injury researchers depend heavily on animal, cadaveric, surrogate and computational models of cord injury.  The selection of model type and its specific design depend largely on the question to be answered.  Both animal and surrogate models were developed in the current work and are the primary focus of this review.  Although small animals (rats, mice, guinea pigs and rabbits) are currently favoured due to economy and availability, this section will predominantly be limited to a discussion of large animal models.  Large animals are considered greater than around 4 kg and include dogs, sheep, pigs, cats and non-human primates (NHPs).  The latter two are frequently low in weight, but cats have relatively large spinal cords for their weight and NHPs are similar to humans in that they are bipedal.  The discussion is also limited to models that achieve SCI via “blunt” mechanical loading and excludes transection models, ischaemic injuries mediated by obstruction of spinal arteries, and administration of chemicals and radiation (reviewed in [353]). 59 1.6.1 Large animal models for experimental SCI Animal models have been developed in a number of animal species and using several injury mechanisms [354,355] each with their own advantages and disadvantages, and therefore suited to address different questions [16,356].  The main motivation for developing in vivo animal models has been to study secondary pathophysiology and assess potential therapies.  For these types of tests, obtaining human-like pathophysiology and measurable recovery from neurological deficit is sufficient, regardless of the mechanical parameters used to produce the injury.  When studying the mechanical phenomena associated with an injury, emphasis should be placed on using mechanical parameters, such as velocity, displacement and loading direction, that replicate human SCI.  If the animal selected has a spine and spinal cord that are similar in size and geometry to human and the research question concerns the mechanics rather than the neurophysiology of the injury, the parameters derived from biomechanical tests on cadavers may be applied directly to the model without need for scaling. SCI research began in large animals; the first report of experimental SCI was in 1911, when Allen devised a method to apply a contusion injury in dogs [357] (see Section 1.6.3.1).  When SCI research intensified in the late 1960s non-human primates were used and remained popular until 1981 [e.g. 294,358,359]; they are now used only sparingly [295,360,361].  Regarding contusion-type models, dogs were used little in the early 1970s, and then consistently from 1975-1982 [e.g. 362,363,364], followed by a single study in 2000 [365].  Cats were used between 1969-1989 and were used almost exclusively in the 1980s [e.g. 1,277,366,367-369]; a sheep contusion model was used by a single group in 1975-1984 [e.g. 218], and by another in 1989 [370,371].  Smaller animals (rats, rabbits, guinea pigs, ferrets and mice) were used sporadically from the 1970s, and rats have been used almost exclusively for the last 20 years. The change to rodent models was largely motivated by societal pressure, financial considerations, and the need for large scale screening of therapies [372].  The studies are too numerous to recount and only selected references have been offered; more detailed histories have been given by Dohrmann [373], Fernandez et al. [354] and Young [267]. In general, the SCI research community is supportive of the development of a large animal model as a pre-clinical testing tool [374].  The pig has wide acceptance as a laboratory animal for medical research [375] including for modeling TBI [376] and iatrogenic spinal cord ischaemia [e.g. 377,378-380]. However, the first report of experimental traumatic SCI in a pig was not until 1990 [380,381], followed by studies by three separate groups in the last five years [382-385].  As detailed in Table 1-11, each has used a different method to impart the injury.  These methods have been used in various large animals previously and are critiqued in Section 1.6.3. Pigs offer several advantages over other large animals and rodents.  They usually present fewer ethical concerns than do domestic pets (such as cats and dogs) and non human primates, and some 60 lineages can be purchased free of common diseases.  There are several miniature breeds available which are beneficial for chronic studies requiring care and handling of the paralysed animal.  Of particular interest to this study is that pigs have similar spinal structure [386,387] and spinal cord vascular supply [388] to humans.  With appropriate selection of breed and age, they can also have similar spinal cord dimensions to humans.  The main disadvantages of pigs relate to lack of knowledge of the spinal cord tract organisation and differences in central pattern generator motor control from humans, although the latter is common to all quadrupeds. Table 1-11  Summary of pig models of traumatic SCI Author Study objective Animal characteristics Model and experimental protocol Owen 1990(A) [380] Effect of distraction duration on neurological deficit measured by evoked potentials, wake-up test and histology. Domestic, M/F* 30-45 kg Injury level: T5/6 Acute Quasi-static distraction. Applied via Kostuik screws (w/ disc osteotomy and PLL transection) until evoked potential lost. Maintained for 3 to 30 minutes. Owen 1990(B) [381] Effect of distraction level and duration on neurological deficit measured by evoked potentials, blood flow, wake-up test and histology. Domestic, M/F 30-45 kg Injury level: T5/6, T12/L1 or L3/4 Acute Quasi-static distraction. Applied via Kostuik screws (w/ disc osteotomy and PLL transection) until evoked potential lost. Maintained for at least 20 minutes. Bernards 2006 [383] Effect of methylprednisolone on intravenous and intrathecal biochemical markers assessed by microdialysis. Farm bred, M/F 18-22 kg Injury level: T13 Acute (to 250 minutes) Weight-drop.  1 cm diameter steel cylinder impounder, aluminum tube guide, 25 g steel weight dropped from 45 cm.  Removed immediately. Zurita 2008 [389] Functional recovery after autologous transplantation of bone marrow stromal cells. Minipigs, F 20 kg Injury level: T12-13 Chronic (to 6 months) Clip compression.  Open dura mater, apply two surgical Heifetz’s aneurysm clips for 20 min.  Suture dura and cover w/ PTFE* sheet. Skinner 2009 [384] Feasibility of EMG* to detect motor tract injury. Young adult pigs Injury level: high thoracic Acute Clip compression (lateral).  Slow or rapid compression to 50% by metal calliper.  Rapid compression held for 14 sec. Zahra 2010 [385] Haemodynamics after complete cervical SCI. Yorkshire domestic, F 5-9 kg, age: 3-5 wks Injury level: C3-4 Acute (to 4 hr) Controlled cortical impactor. (adapted from TBI).  8 mm diameter flat tip.  Complete injury: depth 5 mm, dwell 300 ms, 80 psi. Open dura, drained CSF. Kuluz 2010 [382] Paediatric model development.   Yorkshire domestic, F 5-7 kg, age: 3-5 wks Injury level: T7 Chronic (to 28 days) Controlled cortical impactor. (adapted from TBI).  6 mm diameter flat tip.  Complete injury: 5mm-60psi, 8mm-80psi; Incomplete injury: 3mm-30psi. *M=male, F=female, EMG=electromyography, PTFE=polytetrafluoroethlyene 61 1.6.2 Relative size of animals for SCI models Selecting an animal with spinal cord dimensions similar to human has the potential to improve the biofidelity of the model’s mechanical response.  Three dimensions are of primary importance to the current work: diameter of the spinal cord, diameter of the dural sac (or the thickness of the CSF layer), and the length of the spinal cord.  Physical measurements of the CNS of animal species are sparsely reported in the published literature. The majority of experimental SCIs are performed in the thoracic region, typically between T6 and T12.  This produces hind limb paralysis but with fewer autonomic nervous system complications than higher level injuries, thus reducing the complexity of post-injury animal care [390].  The sagittal and coronal diameters of the human spinal cord are around 4-8 and 7-9 mm, respectively, in the thoracic region [391-394].  In comparison, the thoracic rat spinal cord diameter is around 1.5-2 and 2-3 mm diameter in the sagittal and coronal planes [395], in 3 kg cats the coronal diameter is around 4.5-5 mm and sagittal diameter is around 4.5 mm [396], the medium size dog has a 7 mm coronal diameter [397] and in 22 kg pigs the coronal diameter is around 6 mm [384,398].  Although it can be difficult to obtain a pre-injury measure of cord size, depending on the extent of cord exposure during surgery, surprisingly few studies have reported the size of the spinal cord for the animals they were using.  This makes the comparison of injury severity difficult across different studies that have used different impounder contact areas, different sizes of animals within a species, and different species.  The diameter of the cord is important for biomechanical studies so that the injury parameters can be directly translated from biomechanical cadaver studies of the burst fracture and fracture-dislocation process without requiring relatively uninformed scaling assumptions (see Section 1.6.4).  Also, the importance of uniform spinal cord size within a given study has been highlighted; in a study of experimental SCI in cats, the number of surviving axons, and hence the extent of injury produced by a given impact, varied directly and significantly with the dimensions of the spinal cord [396]. Dural size (or CSF layer thickness) is documented to a limited extent for humans, and is not documented for animals.  Accurate measurement of dura diameter is challenging.  The dural sac changes shape when it contains no CSF (i.e. ex vivo), and in vivo measurements using CT myelography and MR imaging may contain systematic bias due to contrast media concentration, selection of the pulse sequence and window level, partial volume effects, motion artifact and resolution limitations.  From observation, the dural sac of rats is insignificantly larger that the spinal cord, and the dura appears to be separated from the cord by only a thin film of intervening fluid.  In contrast, when the dorsal dural sac of pigs and humans is visualised after surgical exposure, it is clearly separated from the dorsal surface of the spinal cord by a layer of CSF several millimetres thick.  A human-like CSF space is important in biomechanical studies because fluid flow and pressure distribution can be dependent on the dimension of the fluid layer. 62 Further, the fluid layer must be large enough to accommodate the implanted transducer with limited interruption of the natural fluid flow. The adult human spinal cord is around 41-45 cm long [399].  In comparison, the rat’s spinal cord is around 9 cm long [395], adult cats weighing about 2.5 kg have spinal cord lengths of 33.9±1.6 cm (N=12) [400], adult macaca rhesus and macaca irus monkeys have spinal cord lengths of 23.2±3.7 cm and 20.1±1.6 cm, respectively (N=6 each) [401], and in farm-bred pigs weighing 20±1.4 kg the spinal cord length averages 45-55 cm [398].  The length of the spinal cord could be important in the transmission of the energy away from the injury site through the spinal cord, CSF and dura.  It may also contribute to the damping and reflection of such a stress or pressure wave. 1.6.3 Methods of producing injury Since the first documented SCI model by Allen in 1911 [357], a variety of experimental injury methods have emerged.  These can be broadly categorised into dynamic and static models; the former is suitable for modeling traumatic injury, and the latter can simulate either a chronic myelopathy or residual compression resulting from traumatic SCI.  Models are also designated as open: loads are applied directly to the spinal cord and dura after surgical exposure; and, closed: loads are applied to the vertebra and transmitted to the spinal cord via the anatomy.  The following discussion is largely limited to methods that have been applied to large animals, with only a short reference to rodent models.  However, it is noted that rodent models dominate current research, and the SCI devices for rodents are more sophisticated in terms of mechanical control and measurement.  As previously mentioned, animal models need to be assessed in terms of relevant biomechanics, as well as the creation of relevant physiological response.  All blunt injury models exploit the intuitive relation between injury severity and the magnitude and rate of tissue deformation.  The following discussion primarily concerns the biomechanics of the injury methodology. 1.6.3.1 Weight-drop Contusion SCI models in large animals have been based almost exclusively on the so-called “weight-drop” method of Allen [357,402].  Allen is credited with developing the first quantitative and standardised experimental SCI model; this consisted of locally exposing the thoracic spinal cord (dura intact) of the anesthetised dog, placing a tube oriented perpendicularly to the spinal cord, and dropping a weight through the tube.  The severity of the contusion, described in “gram-centimetres”, was modified by adjusting the weight or the height from which it was dropped.  There was little focus on SCI in the following decades and the weight-drop method was only used by Amako in 1935 [403] and Freeman in 1953 and 1963 [404,405].  When SCI research resumed in the late 1960s, the weight-drop model was used profusely with various modifications in dogs, cats and monkeys until around 1985 when it was 63 adapted for the rat [297,406].  There is no generalised or commercially available large animal weight- drop device. Unfortunately in the early use of the weight-drop method, the indeterminacy of the “g-cm” unit was not appreciated.  As described above in Section 1.5.4.1, the impact velocity, force, displacement and impulse have an approximately linear relationship with injury severity (e.g. structural and cellular tissue damage, functional recovery) and mechanical indicators of severity (e.g. deformation and peak force), while impact energy does not.  From elementary mechanics, the potential energy (mgh) of a dropped mass is converted to kinetic energy (½mv2) at impact so that for a plastic collision the energy transferred during the impact is linearly related to both height and mass.  However, the impact velocity is proportional to the square root of the height (since mgh=½mv2; v=2), and so is the impulse (I=m∆v). The impulse defines the peak force and the duration of the impact (I=∫Fdt).  Several authors have shown at length that different weight-height combinations with the same g-cm product do not produce the same mechanical insult or resultant injury severity [276-278,366,367].  Because of the persistent use of the g- cm unit, the impact velocity and mass used in many previous studies cannot be determined, and it is difficult to judge relative injury severity. The specifications of the interface between the device and the animal have a significant bearing on the resultant injury severity and repeatability.  Such specifications include use of an impounder (and its material and mass), and the area, shape and conformity of the impacting tip relative to cord size and shape.  Most devices have used an impounder resting on the dura, either balanced freely, supported by a frame or saddle placed on the vertebra, or residing in the guide tube.  Impounder materials have included Teflon, aluminum, lucite, brass and plexiglass, and masses have varied from 0.1 g up to 14.7 g [367] and 26 g [366] when instrumented with load cells.  The mass of the impounder, particularly relative to the dropped mass, changes the energy transfer between the falling weight, the static impounder and the spinal cord, thereby changing the cord deformation [369,407]; a heavier impounder requires more energy to be accelerated [368] and causes greater overall cord compression during the injury [364].  Loss of energy will also occur due to deformation of the impounder and mass, depending on their relative stiffness. Furthermore, impounders can cause substantial pre-impact dura and cord static deformation.  With an impounder of only 0.1 g the dura deflection was 0.2 mm and with a 10 g impounder it exceeded 2 mm and likely compressed the spinal cord [369].  These results indicate that impounders should be avoided when studying the biomechanical function of the CSF and dura. The contour and area of the impacting tip, whether an impounder or falling weight, has been shown to affect the injury.  Some previous studies used concave contoured impounders [294,358], although flat circular faces were typical.  The diameter of the impacting tip for cat, dog and monkey injuries was generally 5 mm.  A one-fold increase or decrease in cross-sectional area led to a 20-30% inversely 64 proportional change in maximum deformation and peak stress [369].  Molt et al. [277] state that although using a 4, 5 and 6 mm diameter impounder changes the area by around ±50%, no changes in impulse and lesion size were detectable; however, it is not clear that the mass of the impounder was kept constant.  A larger impounder area created by contouring of the tip into an arch which straddled the cord, rather than by increasing the diameter, was found to create a more severe injury and one that better simulated clinical injury [363].  Few weight-drop devices included a mechanism to eliminate “bounce” of the impactor after the initial impact [296]; bounce may lead to secondary impacts which may increase the variability of the induced injury. A variety of materials have been used for the guide tube (e.g. Teflon, glass, plexiglass, brass, stainless steel) and the falling weight (brass, lead, steel, mercury-filled glass tube).  The selection of weight-guide pairs would affect the friction characteristics, as would the configuration of holes in the guide tube and the size of the opening at the spinal cord end.  High friction and air resistance would reduce the impact velocity; however, none of the early models report impact velocity.  The original design by Allen used an open frame rather than tube [357], one device used a central shaft [358] and one relied on accurate alignment of the weight above the impounder and omitted a guiding device [1].  The guide tube mount is not described in most studies; for some it was attached to the surgical table [368] or a stereotactic frame [408],  while others rested on saddle which sat directly on the vertebra [294].  One author reports inducing unilateral pneumothorax to reduce body movement due to chest expansion during respiration [369], but others do not report how thorax motion was compensated for in determining drop height for table-mounted devices.  In contemporary rat models chest excursion is generally eliminated by partial suspension of the body weight on vertebral clamps attached to a stereotaxic frame. The vertebral clamps also serve to stabilise the vertebral column.  The need to eliminate motion and flexibility of the spinal column and ribcage under the point of injury was identified by some early researchers.  In the rat, this is usually achieved by restricting the laminectomy to a single level and clamping the spinous processes of the adjacent vertebra to the stereotaxic frame [407].  However, vertebral clamping and weight suspension are more difficult to achieve in larger animals, and there are few reports that it was considered.  In one device a metal saddle rigidly fixed to the injury apparatus was inserted under the cat’s spinal cord [296], and in another, rigid supports were placed under the transverse processes and in turn attached to the surgical table [396]. Measurement of the mechanical input and output parameters, such as velocity, displacement and force, was largely limited to studies which focused on the biomechanics of the model and were not reported by studies of SCI pathophysiology.  Two groups incorporated a load cell into the impounder [358,366] for exploring mechanics, but this device was not used to describe injuries for actual pathology or therapeutics studies.  To account for the chest and spine deflection noted above, deformation of the 65 spinal cord must be measured independently of chest excursion [366], although not many early models collected these data.  In most experimental models the rebound of the dropped weight was assumed insignificant; the rebound elimination mechanism for one apparatus was a string attached to the weight which was pulled manually after impact [409] and another using an automatically triggered electromagnet to stop the weight in its upward flight [296]. In addition to the device-related differences described thus far, variation in biomechanics and pathophysiological response can also be due to animal selection and surgical technique.  Some authors reported wide ranges of animal weights or used different breeds within one study.  Spinal cord dimensions were rarely stated, but variation in size would have altered the load distribution and the relative cord compression, thus contributing to inconsistent injury severity [410].  Some studies opened the dura prior to injury [411,412], which would remove the mechanical contribution of CSF and dura, but the majority did not.  Other factors which may cause variation include the tension in the dura and cord, tissue and fluid pressure, blood pressure, relative diameter of cord, subarachnoid space and canal, and proximity of the injury site to nerve roots and dentate ligaments [407]. The only standard weight-drop device is for rats: the New York University (NYU) impactor [413] also referred to as the MASCIS (Multicenter Animal Spinal Cord Injury Study) impactor as it was adopted for that large study [355].  Biomechanical standardisations put in place to increase repeatability across the MASCIS institutions included: impactor mass and tip size and shape, height settings for graded injuries at a specified spinal level, and support of the vertebral column [355].  Non-mechanical standardisations  included rat strain and age, anaesthetic protocol, and timing of the injury after anaesthesia induction [267].  The device can measure impact velocity, compression and rate, and force. 1.6.3.2 Controlled displacement and controlled force contusions A number of machines have been designed that produce injury while controlling either peak force or rate/peak displacement.  These machines have been critical in determining the injury response to mechanical parameters that cannot be independently manipulated using a weight-drop device.  These devices control the impactor’s trajectory for both the loading and unloading path, thereby eliminating secondary impacts due to bounce that can occur with weight-drop devices. Although these machines have distinct advantages in the standardisation of injuries, none have been implemented for animals larger than a rat or ferret.  There are currently two standard methods for delivering a controlled contusion injury to rats: the Ohio State University (OSU) impactor and the Infinite Horizons (IH) device [356].  The OSU impactor is a displacement-controlled electromechanical actuator [414-418].  Although capable of delivering speeds of 0.3 m/s, it is typically operated in the range of 0.08-0.1 m/s, to a depth of 66 0.8-1.1 mm (around 30-50% of cord diameter), with a dwell of 4-5 ms at maximum displacement and total event duration of about 10 ms [355].  Because it is displacement controlled, the surface of the dura must be touched to obtain the “zero” displacement before the test is started; some researchers believe that this has the potential to damage the spinal cord.  Displacement and force trajectories are measured, as well as the vertebral column deflection [267].  Displacement control is a very stable and repeatable control mode.  The IH impactor is a force-feedback stepper motor device, typically operated to obtain a peak load of 1-2.5 N (100-250 kdyn) and has a peak speed of 0.13 m/s [419].  No pre-injury displacement “zeroing” is required.  Force-feedback is a challenging control mode, particularly at high rates and when compressing non-linear viscoelastic materials, so the repeatability can be less than displacement control. For both the OSU and IH impactors, active control of the impactor tip eliminates the impactor rebound that can occur for weight-drop devices. Of note is the device of Anderson and colleagues [420] in which contusion was applied to the ferret spinal cord through the intervertebral foramen, without a laminectomy.  This was a pneumatic device in which the maximum stroke was controlled by mechanical stopper.  It was capable of velocities ranging 0.5 to 10 m/s, far greater than the above electromagnetic actuator devices [266], but it lacked a force transducer.  The controlled cortical impactor (CCI) is a similar device; originally designed to impart focal brain injury, versions of this device have recently been employed to induce thoracic [382] and cervical [385] SCI in piglets, as mentioned in Section 1.6.1, and thoracic SCI in rabbits [196]. 1.6.3.3 Vertebral distraction and fracture-dislocation Few models of closed-column (i.e. no laminectomy) dislocation have been reported in the literature, despite fracture-dislocation being one of the most common causes of SCI.  The first was described only briefly but appears to have achieved a dorsoventral dislocation via weight-drop onto, or between, the T9-T11 spinous processes of the cat [421].  As discussed in Section 1.5.4.2, a closed column dog model was used to study the CSF pressure response to lateral impact.  A preset displacement was applied to the intact lumbar spine using a 4 cm diameter captive piston driven by a large free-falling weight.  This method achieved spinal fractures in two, and neural damage in five, of six animals [3].  A pneumatic device to dislocate the L1/2 vertebrae of the dog is described by Fialho [362]; after dissection of the spinous processes and facet joint of L1, the L1 vertebra is held stationary and L2 is displaced laterally.  Both of the contemporary dislocation models were designed for rats: the first holds T12 stationary, leaves T13 unconstrained and displaces L1 laterally to a set displacement at up to 0.15 m/s [422]; the second imparts a dorso-ventral dislocation of specified displacement at C4/5, at speeds of up to 1 m/s [423]. Vertebral distraction models have provided low-rate loading of the spinal column by fixing one of the pelvis/torso and head, and applying a tensile load to the other, in non-human primates [300] and cats 67 [299,424].  In the monkey [425], cat [426] and pig [380,381], distraction was achieved by resecting an intervertebral disc and loading the adjacent vertebra in tension via screws or clamps attached to the spinous processes or vertebral bodies.  Early computer controlled distraction models used a materials testing machine, attaching adhesive clamps directly to the spinal cord (dura removed, pia intact) in vivo in cats [219] and dogs [221,241], and measured motor and sensory function for up to three weeks post- injury.  Most recently, computer controlled vertebral distraction has been achieved in the rat at speeds of up to 0.9 m/s via laminar hooks at T9 and T11 [427], and at 1 m/s via dual-vertebra clamps placed across C3/4 and C5/6 [423]. 1.6.3.4 Residual compression models Although the aforementioned models all have the potential to provide a prolonged compression after the dynamic injury, they are generally used in a purely acute manner.  Dwell times for the electromechanical devices are generally 4-5 ms [355], while weight-drop and other non-computer controlled methods have dwell times on the order of seconds because the compression is manually removed or released.  A number of models have been designed to emulate the residual compression commonly present after traumatic SCI.  The duration of static compression has ranged from 1 minute to days or weeks.  Some of these devices are applied to the spinal cord dynamically but without rate control. Furlan et al. [200] provide a comprehensive review of residual compression models that have been used to study surgical decompression. The clip compression model uses an aneurysm clip to apply a constant, dorsoventral and extradural force to the spinal cord for a specified length of time [428].  The closing force of the spring-loaded clip is calibrated in grams, and values corresponding to mild, moderate and severe injuries in the cervical and thoracic spinal cord have been established for the rat [429].  Although the jaws of the clip are released quickly, the dynamic parameters of the compression are not measured, and the calibrated force applies to the static condition only. Other residual compression methods have used pistons fixed to the vertebrae (0.17 m/min then static) [430,431], ligature-type cables [208,209], and screws protruding through the vertebral body [432- 434] in dogs and cats [435], and spacers hooked under laminae adjacent to a laminectomy in rats [207,436].  Modified forceps have also been used to create a predetermined static compression [410]. Weights have been “rested” on the exposed thoracic dura of cats for different durations including: 18-58 g for up to 20 min [437,438], 200g/5min, 600g/15min [439], 170g/5min [440,441].  Similar tests have been done in ferrets [442], dogs (6-60 g for 30-60 min) [443], monkeys (50g for 2 hours) [360] and rats (20/35/50 g for 1/5/10 min) [444]. 68 The epidural balloon catheter, first used in the dog by Tarlov [292,445], has the advantage of not exposing the spinal cord at the injury level.  A balloon-tipped catheter is introduced via a laminectomy at an adjacent vertebral level, and advanced to the injury level in the dorsal epidural space.  The balloon is then inflated to a specified pressure and  is assumed to result in a specified spinal cord compression; however, few authors have verified this and it is a potential source of variation in the method.  One study modified this approach by using an inflatable cuff that was placed around the circumference of the cord after a laminectomy at the injury level [216].  Some authors report a “rapid” inflation and immediate (within 30 sec) deflation of the balloon [446,447], while others use it to compress the cord for periods greater than one minute and up to several weeks [447-450].  Regardless, the balloon catheter is not considered a “dynamic” device, compared to the other contusion models where the impact has a duration of 5 to 20 ms.  The balloon catheter method has been used in the dog [365,447,451-454], cat [446,455], monkey [448-450,456-459] and lamb [370,371]. 1.6.3.5 Comparison of the injury strategies Each injury model has strengths and limitations which makes it suited to studying particular aspects of the biomechanics of the injury.  Large animal models present some unique challenges in device design.  The requirements for sterility are more stringent than for rodents, particularly for chronic protocols or those that require long periods of anaesthesia.  This can influence the type of device selected and aspects of its design such as materials and transducer selection. For studying the physical behaviour of the spinal cord-dura-CSF during the injury, the weight-drop method is probably the most representative of the typical non-penetrating human injury.  No control or limit is placed on the peak displacement or load so that the dura, CSF and spinal cord are compressed until the applied energy is absorbed by the tissues, in much the same way that it can be imagined a bone fragment comes to rest in the spinal canal.  Open contusion injuries are intended to replicate the anterior- posterior compression seen in the burst fracture and fracture-dislocation mechanisms of injury.  However, the contusion is applied to the dorsal surface while the ventral surface is more likely contacted in a human burst fracture SCI.  Closed column dislocation and distraction models could conceivably better replicate the bone-cord interaction, but in practice they lack repeatability, do not directly measure spinal cord loading and deformation, and for large animals the externally applied forces must be very large to disrupt the spinal column.  They can also produce an unstable column, which may cause complications in a chronic protocol. Models that measure load and displacement have the advantage that animals can be eliminated from the study if the biomechanical parameters of their injury lie outside pre-defined acceptable corridors [268,271].  Stabilisation of the spinal column has been identified as a challenge to all open models and is 69 probably more challenging to address for large animals due to their higher weight and greater respiratory motion. A summary of advantages and disadvantages of the SCI models is presented in Table 1-12. 70 Table 1-12  Summary of types of experimental SCI models, the animals used, and the advantages and disadvantages of each. Injury Model Animals Advantages Disadvantages Contusion Weight-drop (open) Dog, Cat, NHP, Rabbit, Rat, Mouse, Sheep, Pig • Similar to real injury • Unrestricted velocity • Dorsal injury • Laminectomy required • Uncontrolled unloading (bounce of weight) • Force, velocity, displacement not usually measured EM or pneumatic actuator (displacement control) Rat, Ferret, Piglet • Controlled unloading • Measure force and displacement • Dorsal injury • Laminectomy required • Relatively low velocity • Establish zero position EM actuator (force control) Rat • Controlled unloading • Measure force and displacement • Dorsal • Laminectomy required • Relatively low velocity Distraction Directly coupled to spinal cord or dura Dog, Cat  • Real injury mechanism • May replicate canal-cord contact • Difficult to prevent grip slippage Via bony anatomy NHP, Cat, Pig, Rat  • Real injury mechanism • May replicate canal-cord contact • Lacks repeatability • More complex machine Dislocation/Fracture (closed) Vertebral weight drop Dog, Cat • Potentially more biofidelic • Canal intact at injury site • Reduced pre-injury surgery • Very large weight/height required • Lacks repeatability • Force, deformation of spinal cord not measured Controlled vertebral displacement Dog, Rat • Potentially more biofidelic • Canal intact at injury site • More complex machine required • Vertebral clamping is critical • Unstable spine post-injury Quasi-static compression Clip compression Rat, Guinea pig • Simple equipment • Dorsal-ventral • Rate of application uncontrolled • Uninstrumented for force, compression rate • Force calibration may change over time Resting weight  Dog, Cat, NHP • Simple  • Rate of application uncontrolled • Uninstrumented for force compression rate Spacer/bead/ligature Dog, Cat, NHP, Sheep • Canal intact at injury site  • Rate of application uncontrolled • Deformation not measured Epidural balloon Dog, Rat • Canal intact at injury site • Rate of application uncontrolled • Deformation not measured, full expansion assumed   71 1.6.4 Selection of mechanical input parameters A SCI model that seeks to ellicit a biofidelic mechanical response and obtain a human-like SCI must faithfully simulate the mechanical inputs to the injury event.  These mechanical inputs may include load magnitude and rate, displacement and velocity.  Closed models seek to simulate the impact between the impacting object and the external anatomy.  This loading regime can be estimated from accident reconstruction methods.  Open models seek to simulate the impact between two or more parts of the internal anatomy, for example a bone fragment contacting the spinal cord during a burst fracture.  The appropriate mechanical inputs for this scenario are more difficult to estimate. Depending on the selected method of load application (see Section 1.6.3), an experimental contusion SCI could be specified with mechanical parameters such as impact velocity, maximum displacement or load, rate of application of the displacement or load, or applied energy.  To model a burst fracture injury mechanism, the ideal open weight-drop model would use an object with the same mass and cross-sectional area as a typical bone fragment, and an impact velocity matching that with which the bone fragment is retropulsed into the spinal canal.  However, determining these parameters is not trivial. Two in vitro biomechanical studies have measured the rate of canal occlusion while inducing burst fractures by applying a dynamic axial load to thoracolumbar spinal segments with the spinal cords removed.  Panjabi et al. [132] produced 13 burst fractures in 9 human specimens with an incremental trauma method, and provided example occlusion-time data for T12 and L1 for one specimen (Figure 1-25).  From this data, it is seen that after an initial 1 mm of relatively low-rate occlusion, the velocity was approximately linear until the peak occlusion, and this occurred over 2.25-2.5 ms.  The occlusion velocities for this specimen were determined to be 2.4 and 1.6 m/s for T12 and L1 respectively.  72  Figure 1-25  Canal occlusion (mm) and compression force (kN) during an experimental burst fracture in the thoracolumbar spine.  Example data of canal occlusion versus time at the T12 and L1 vertebral levels for specimen SP7.  The red dashed lines indicate the start and end points used for calculation of the velocity.  The velocities of the occlusions were approximated by the slope of the blue lines.  Adapted from Journal of Spinal Disorders, 8(1), Panjabi et al., Dynamic canal encroachment during thoracolumbar burst fractures, 39-48, 1995, with permission from Lippincott Williams and Wilkins [132].  The authors provided the maximum occlusion for all the other burst fractures in the series, and the mean velocities were derived for these, between 0.75 mm and peak occlusion and assuming a duration of 2.25 ms, as per Table 1-13 below.  The mean occlusion velocity was 2.1±0.8 m/s. Table 1-13  Canal occlusion velocities for burst fractures at T12 and L1.  Derived from data of Panjabi et al., 1995 [132]. Level T12 Level L1 Specimen # Occlusion (mm) Velocity (m/s) Specimen # Occlusion (mm) Velocity (m/s) SP1 5.1 1.9 SP3 5.4 2.1 SP2 3.4 1.2 SP4 5.3 2.0 SP3 7.4 3.0 SP5 5.7 2.2 SP7 8.6 3.5 SP6 4.7 1.8 SP8 4.6 1.7 SP7 7.0 2.8 SP9 2.6 0.8  By applying high velocity axial loads in a drop-tower apparatus, Wilcox and colleagues [134,460] produced burst fractures in three-level thoracolumbar bovine specimens (spinal cord and dura removed) and visualised the transient canal occlusion with an arrangement of high speed cameras and mirrors. Although bone fragment velocities are not reported directly, they can be estimated from the representative data presented graphically as percent canal occlusion versus time (e.g. Figure 1-26, left).  This, in 73 combination with scaled photographs of representative specimens from which the estimated anterior- posterior canal dimension ranged from 11 to 15.7 mm (Figure 1-26, right), results in bone fragment velocity estimates of 0.95 – 3.52 m/s.  Finite element simulations [133,461] and in vitro simulations [462] derived from this work have used bone fragment velocities ranging 1 – 10 m/s and 2.5 – 7.5 m/s, respectively.  Figure 1-26  Percent spinal canal occlusion versus time for a representative in vitro experimental burst fracture (left)[134][134][134][134][134]; Photograph of representative bovine vertebra for estimating anterior-posterior canal diameter (right). (left) Reproduced from Journal of Biomechanics, 35(3), Wilcox et al., Measurement of canal occlusion during the thoracolumbar burst fracture process, 381-84, 2002, with permission from Elsevier [134]. (right) Adapted with permission from Wilcox, A biomechanical investigation of the burst fracture process, 2002, PhD Thesis, University of Leeds, ©Ruth Wilcox [460].  Ivancic et al. [463] simulated bilateral facet dislocation in cervical functional spinal units using an impact sled and measured the dynamic canal diameter.  The mean peak canal narrowing velocity was 0.23 ± 0.06 m/s for 10 specimens, which is substantially lower than the values obtained in the burst fracture studies. 74 The mass and cross-sectional area of bone fragments arising from the aforementioned in vitro burst fracture investigations were not reported in these biomechanical studies.  However, subsequent work from Wilcox and colleagues [462] reports a design criterion for synthetic bone fragments in the range of 5.8 ± 1.7 g based on fragments collected from the bovine specimens aged 14-21 days [460].  These vertebral bodies had a lateral width and anterior-posterior depth (e.g. Figure 1-26, right) comparable to human thoracolumbar vertebrae (35-50 mm and 30-35 mm, respectively [464]).  As shown below, this is lower than the impactor weights used for large animal weight-drop models.  In comparison, the wet weight of human thoracolumbar vertebrae ranges from 250-550 g [465]. To the author’s knowledge, none of the previously used large animal models of SCI have utilised mechanical parameters derived from an analysis of human injury.  In fact there is little justification given for the various combinations of height (velocity) and mass that have been used in the various weight drop apparatus.  Graded models typically utilised empirically determined parameters to produce the desired injury severities based on survivability, short- or long-term functional recovery and lesion distribution and shape.  These are arguably satisfactory metrics for the study of injury physiology, but probably not ideal for biomechanical studies.  Figure 1-27 illustrates the drop heights and weights that have been used in previous large animal models, and the corridor for human burst fracture parameters as indicated above (dotted black rectangle).  This shows that the majority of large animal models have used higher masses and lower velocities than have been measured for burst fracture type mechanisms in in vitro spine specimens. 75  Figure 1-27  Injury parameters reported to induce transient or permanent paresis in various large animal weight-drop models. The data was collated from all available large animal studies that used an open weight-drop model and reported separate drop weight and height information, from 1968 to present. The weight-height combinations used in the studies presented in Chapter 2 and 3 are also shown (filled red markers).  Dashed areas indicate nominal “injury corridors” for cat (orange), dog (dark red) and monkey (green) models.  Dotted black rectangle indicates the range of heights that would theoretically produce occlusion velocities measured for burst fracture and dislocation events observed in the biomechanical experiments described in the text above.  1.6.5 Synthetic models of SCI Several synthetic models have been devised to study the mechanics of blunt spinal cord trauma. Some have included separate elements for spinal cord, CSF and dura, with varying effort to accurately replicate the mechanical and structural properties of the synthetic “tissues”.  Synthetic spinal cords are advantageous due to the rapid changes in mechanical properties of the natural spinal cord after death [227].  Synthetic spinal cords have been fabricated from Tygon tubing [466,467], gelatine [468,469], silicone elastomer [280,470], and silicone gel with Dacron fibres [471,472].  Dura does not degrade rapidly if adequately hydrated and may be stored frozen, however few studies have incorporated natural dura or a surrogate material.  One study fitted excised bovine dura over synthetic cords [280], and one study used a vinyl rubber glove material [281] to simulate dura in a bench-top model. 76 Hall and colleagues [280,281,462] simulated burst fracture spinal cord impingement by projecting a simulated bone fragment at bovine and/or synthetic spinal cords suspended in a materials testing machine and recorded the resultant deformation.  These models showed that decreased bone fragment velocity, presence of and increased tension in the posterior longitudinal ligament (PLL) [462], presence of CSF [280,281] and increased bone fragment cross-sectional area [281] reduced the peak deformation of the spinal cord, but dura alone did not [280,462].  However, in these studies the cord was not directly visualised, and an assumption of full collapse of the subarachnoid space was made in order to derive maximum spinal cord compression. Hall et al. [462] reported peak pressures in the spinal cord of around 300-560, 600-2250 and 1425- 2700 mmHg at the injury site, with impact velocities of 2.5, 5 and 7 m/s, respectively, with different PLL tensions and posterior element configurations.  In a similar model that used a weight-drop rather than projectile impact, Pintar [468] and Yoganandon [469] placed a synthetic spinal cord instrumented with squares of pressure sensitive film into cadaveric cervical spines with a C4 laminectomy, and subjected them to weight-drops of 100 to 600 g-cm.  Peak pressures ranged approximately 1425-3075 mmHg at the impact site, and 225-450 mmHg at the adjacent spinal level. Some studies have simulated loads to the spinal column to induce fracture and measured the response of the spinal cord to canal contact.  The aforementioned instrumented synthetic cord was placed in two cadaveric cervical spines, with heads, that were subjected to axial loading at 3.1 and 6.1 m/s [469]. The first recorded a peak pressure of 1507 and 188 mmHg corresponding to a C5 burst fracture and C4 wedge fracture, respectively.  The second recorded peaks of 83 and 675 mmHg, adjacent to a mild C5 compression fracture and a C3 compression fracture, respectively. Xie et al. [473] removed the spinal cord of eight thoracolumbar cadaver specimens, placed a pressure transducer against the posterior canal wall at L1 and filled the dura with water, sealing the intervertebral foramen.  The specimens were potted to leave L2-4 free, and subjected to a 5.4 m/s axial impact by weight-drop.  They observed positive pressure waves of 217-2415 mmHg associated with burst fractures.  When no burst fracture occurred, there were both positive and negative pressure waves, with a range of 488-607 mmHg, which they attributed to intervertebral disc deformation. Chang [466] and Tran [467] utilised an occlusion transducer that measured changes in pressure in a fluid filled tube placed in the canal of spines subjected to axial loading.  They do not report pressures except for one example trace which shows a clipped signal at 250 mmHg during a C3 compression fracture. Bilston et al. [472] subjected a physical head-spine model to rapid hyperflexion/extension and measured strains in a surrogate spinal cord with an internal ink grid.  In flexion, strains were highest in 77 the cranial cord, at 20-35%, with strain rates 7-19 s-1.  In extension the cord tended to shorten locally, tensile strains were not above 12%, and strain rates ranged 1-12 s-1.  The same model was used in simulations of head-first impact, where flexion-compression loading caused C2 dislocation and extension-compression loading caused C4/5 dislocation [474].  Adjacent to the injury location, cord strains were up to 40%, and strain rates were up to 6.5 s-1. Only some of these studies report considering the application of an appropriate spinal cord and/or dura pre-strain.  These are likely to contribute substantially to the mechanical response of the model during the impact. 1.6.6 Summary The complex mechanical and physiologic aspects of SCI necessitate the use of animal, synthetic and computational models that attempt to replicate various aspects of the injury event.  Contemporary large animal models of SCI are sparse, despite offering unique features of human-like scale and physiologic similarity, both of which are advantageous for pre-clinical and biomechanical studies. A variety of contusion, dislocation, distraction and static compression models of SCI have been used over the last five decades.  Devices that provide a controlled displacement or load profile are useful for producing consistently graded injuries and studying the independent effects of mechanical variables. However, the weight-drop method produces a time-varying rate of compression, and has the potential to best mimic the biomechanics of a contusion SCI such as that resulting from a burst fracture or fracture- dislocation.  When applied to an animal with a spinal cord of similar dimensions to humans, appropriate injury parameters (impact velocity and mass) may be estimated from in vitro biomechanical tests which have induced common spinal fracture types and measured the velocity and mass of the bone fragments and/or canal occlusion produced.  In vivo large animal models currently provide the most biofidelic simulation of the human spinal cord/dura/CSF response to mechanical insult. Several synthetic and cadaver models have also been used to study the effect of various anatomical features, such as the dura, PLL, canal width and spinal degeneration, on spinal cord deformation and pressure.  The advantages of such models include control of physical dimensions and ease of test repetition and elimination of physiologic effects; disadvantages include the simplification of the anatomical structure and material properties of the system.  These studies have shown to some extent that tissue and fluid pressure pulses occur during impacts that simulate SCI.  However, few have incorporated dura and CSF into the model.  78 1.7 Measuring CSF pressure in the brain and spine The salient characteristics of normal CSF pressure were introduced in Section 1.3.3, and pressure transients in the CSF and parenchyma during experimental CNS trauma were discussed in Sections 1.5.4.2 and 1.5.4.4.  This section reviews and critiques the measurement locations and transducer types that are, or have been, used for quasi-static measurement in clinical settings and for dynamic measurements during experimental CNS trauma. 1.7.1 Quasi-static clinical and experimental pressure measurement CSF pressure is routinely measured in the brain and to a lesser extent in the lumbar spine.  It is used for the clinical diagnosis and monitoring of pathologies such as hydrocephalus and traumatic brain injury.  Spinal CSF pressure is measured in the lumbar cisterna where the existence of the cauda equina provides an ideal location to advance a needle and catheter into the subarachnoid space with minimal risk of neural damage.  Cranial CSF pressure, or intracranial pressure (ICP), is measured in the ventricles or parenchyma and less commonly in the subarachnoid and epidural spaces.  Although early experimental studies used water or mercury manometers, clinical and experimental transducers are now largely based on strain gauge and fibre optic technology. The gold standard for ICP and lumbar CSFP monitoring is an intraventricular or lumbar catheter and rigid-walled tubing connected to an external pressure sensor, usually with a semiconductor strain gauge transducer [54,475].  The catheter may also function as a drain to remove fluid.  The external transducer must be levelled to a standard anatomical reference position, most commonly the midpoint of the right atrium [80]; it is periodically exposed to atmosphere, via a stopcock, for zeroing. Both fluid and parenchymal measurements are performed with indwelling “microtip” transducers. Those currently available are based on semiconductor strain gauges (e.g. Codman, Johnson and Johnson, Raynham, MA; Neurovent-P, Raumedic AG, Munchberg, Germany) and fibre-optic cavities (e.g. Camino, Integra Neuroscience, Plainsboro, NJ).  The primary concerns with indwelling transducers are zero drift and temperature sensitivity, because they cannot be zeroed in situ.  Drift has been evaluated for a number of clinical ICP transducers using bench testing [476-480] and in clinical practice [481-483], although the latter is challenging given that minor atmospheric changes are significant in physiologic pressure terms. Clinical transducers are typically unsuitable for dynamic measurements because they have a low frequency response.  Further, the diameters of the catheters or indwelling transducers used for clinical applications are around 1 to 3 mm, which is relatively large compared to the spinal CSF thickness.  These limitations are addressed by some of the transducers introduced below. 79 1.7.2 Dynamic experimental CNS injury pressure measurement Transient CSF pressure has been measured in the spine in only three studies (see Section 1.5.4.2). This discussion is therefore broadened to include transducers used for cranial CSF and parenchymal pressure measurements, some of which were introduced in Section 1.5.4.4.  Most researchers have recognised the importance of using transducers with a sufficiently high frequency response.  These transducers can be grouped into three categories: piezoelectric, piezoresistive and fibre optic. Piezoelectric transducers utilise a piezoelectric crystal which responds to mechanical strain by generating a voltage.  They have inherently high natural frequency and exhibit linearity over a wide range, but can only be used to measure dynamic events.  They typically have a relatively large metal casing, which means they need to be surface-mounted to a rigid interface, such as the skull.  Due to this casing they are considered a robust device.  Early studies of the pressure transmitted to the brain from explosive detonation used piezoelectric transducers mounted in the orbits of deceased rabbits [484] and monkeys [485].  Others have mounted similar types of transducers through trephined holes in the skull of cats undergoing fluid percussion TBI [313] and swine adjacent to firearm noise [342].  These authors report puncturing the dura but not sealing it around the transducer.  The spinal CSF pressures measured in the closed weight drop dog model described earlier were made with a piezoelectric transducer coupled to a fluid-filled catheter [3]. Piezoresistive transducers consist of semiconductor or metal strain gauges bonded to a flexible membrane.  The strain gauge changes resistance in proportion to the deformation of the membrane which is caused by the incident fluid pressure.  These transducers have a reasonably high frequency response and can be used for both static and dynamic measurements.  Piezoresistive transducers have been skull mounted in post mortem human subjects (PMHS) during frontal head impact experiments [486] and in swine subjected to whiplash motions [288,487].  Miniaturised piezoresistive transducers have been mounted at the end of flexible catheters to allow implantation into the brain parenchyma of swine [319,488] to measure the response to fluid percussion and blast TBI, and into the spinal subarachnoid space of PMHS [289] and swine [287] in whiplash simulations.  The piezotransistor or “Pitran” transducer was coupled to a fluid-filled catheter to measure the spinal CSF pressures adjacent to a weight drop SCI in cats [1,2]. All three of the studies that have measured spinal CSF pressure during experimental SCI used transducers that were coupled to the subarachnoid space via some length of fluid-filled catheter and tubing [1-3].  This method is not ideal because, depending on the length of the tubing, it can lead to considerable damping of the signal magnitude and stretching of the signal duration.  This effect was explored by Wennerstrand et al. [3] using a bench-top model in which the signal from a reference 80 transducer placed in direct contact with the fluid was compared with the fluid-filled catheter transducer placed the same distance from the impact (Figure 1-28).  Figure 1-28  Comparing a fluid-filled catheter transducer and indwelling transducer: apparatus schematic and results.  (left) Schematic of a physical model used to test a fluid-filled catheter pressure transducer (as used in animal experiments) against a reference transducer in direct contact with the fluid. (right) Simultaneous responses of the two pressure transducer systems.  Curve 1: fluid-filled catheter system, Curve 2: reference transducer, Curve 3: magnification of curve 1, Curve 4: magnification of curve 2.  Adapted and reproduced from Journal of Biomechanics, 11(6-7), Wennerstrand et al., Mechanical and histological effects of transverse impact on the canine spinal cord, 315-31, 1978, with permission from Elsevier [3].  Fibre optic transducers are a contemporary technology that utilise the Fabry-Perot optical interferometry principle.  The sensor tip contains a very small optical cavity between two partially reflective parallel surfaces.  The outer surface is a diaphragm that deflects under pressure, thus causing a change in the optical cavity depth, and a subsequent change in the cavity reflectance.  When a light signal is transmitted down the fibre to the sensor, the intensity of the reflected light is proportional to the pressure acting on the diaphragm.  These transducers have the advantage of being extremely small and flexible, which is advantageous for implanting in small fluid spaces to enable direct and highly localised contact with the fluid or tissue.  They do not need to be mounted rigidly to bone; rather they can be attached to soft tissues using adhesive or suture.  Their low mass and flexibility reduces the extent to which they alter the mechanics of the system.  However, they are delicate and therefore prone to breakage, and can be cost prohibitive.  Miniature fibre optic transducers have been used in a rat to measure ventricular CSF pressure response to blast overpressure [337], and in rabbit brain parenchyma to measure response to fluid percussion, controlled cortical impact and weight-drop experimental TBI [314]. 81 Pressure changes in the brain parenchyma of rabbits due to rotational acceleration impulses have also been measured with fibre optic transducers [290]. The characteristics of the transducers that have been used for dynamic measurements of pressure associated with CNS trauma are detailed in Table 1-14. 82 Table 1-14  Comparison of pressure transducers used to characterise dynamic experimental CNS injuries, grouped by type. Note that where specifications were not stated in the paper, attempts were made to find details from the manufacturer. Author Location [animal] Model / Manufacturer Size (mm) Range / Linearity (mmHg relative) Frequency characteristics Piezotransistor Albin 1975 [2] Hung 1975 [1] Thoracic spine CSF, via fluid filled catheter [cat] Pitran P-10 / Stow Laboratories NR NR  / 0.5% Natural frequency 150 kHz Piezoelectric Clemedson 1956 [484] Romba 1961 [485] Cranial CSF, orbit mounted [rabbit, monkey] BC-10 / Atlantic Research NR 0 – 15515 / to 517150 / ±1% acc.   Frequency range 3 Hz-75 kHz Wennerstrand 1978 [3] Thoracic spine CSF, via fluid- filled catheter [dog] LC-10 / Celescon Industries Inc. NR 75 – 75000 / NR Frequency range 0.2 Hz-100 kHz Rinder 1968 [489] Stalhammar 1987 [313] Brain, flush with dura through trephined hole [rabbit, cat] Type 6-01 / Vibrometer NR NR Natural frequency 150 kHz; dynamic rise time <0.1ms Saljo 2008 [342] Brain parenchyma, skull mounted, dura punctured [swine] Model 8103 (hydrophone) / Bruel & Kjaer 9.5 mm dia. NR Frequency range 0.1 Hz-20 kHz Piezoresistive (semiconductor or metal foil strain gauge) Sullivan 1976 [315] Brain, flush with dura through trephined hole [cat] PA 85-100 / Statham Laboratories NR NR Natural frequency 10 kHz Nahum 1977 [486] Ortengren 1996 [487] Svensson 1993 [288] Surface flush with internal skull, opened dura at PT site [PMHS, swine] 8510-100 / Endevco  5 mm dia.; (3.8 mm) face diameter -750 – 5250  / NR Natural frequency 180 kHz Frequency range 0-10 kHz  Suneson 1990b [490] Brain parenchyma, skull mounted, dura punctured [swine] 8514-100 / Endevco mounted in thin polyethylene catheter NR 0 – 5171 / NR NR Walter 1999 [319] Brain parenchyma [swine] MikroTip  / Millar  1.5 mm dia. ± 300 / NR 100 kHz freq.  response (-3dB)  83 Author Location [animal] Model / Manufacturer Size (mm) Range / Linearity (mmHg relative) Frequency characteristics Eichberger 2000 [289] Boström 1996 [287] Cranial and cervical spine subarachnoid space [PMHS] Cervical spine, subarachnoid space [swine] Catheter tip / NR NR NR NR Fibre optic (Fabry-Perot optical interferometry) Chavko 2007 [337] Ventricle via skull trephine hole [rat] FOP-MIV / Fiso Technologies Dia.  0.55 mm sensor, 0.25 mm fibre w/ 0.9 mm coating 0 – 400 / NR Sampling rate 100 kHz Bauman 2009 [338] Brain parenchyma and ventricle [swine] FOP-MIV / Fiso Technologies NR NR NR Clausen 2005 [314] Ventricle via skull trephine hole [rat] NR / Samba 0.34 mm dia. 684 – 3800 / NR Sampling rate 500 Hz Krave 2005 [290] Brain parenchyma [rabbit] Samba 3000 / Samba Sensors AB 0.42 mm sensor, radiopaque coated fibre 0.45 mm dia. Enclose fibre in 0.9mm OD PTFE tube -535 – 3740 / NR Sampling rate 15 kHz *PMHS = post mortem human subject NR=not reported 84 1.7.3 Summary Cranial and spinal CSF pressure measurement is practiced clinically as a diagnostic, prognostic and monitoring tool.  Quasi-static measurements are usually done with fluid-filled catheters connected to external piezoresistive pressure transducers, or miniature indwelling piezoresistive or fibre-optic devices. A primary concern for long-term CSF pressure measurements is transducer zero drift, and for external transducers strict attention must be paid to the zeroing protocol to achieve accurate physiologic values. The response of CSF pressure to an experimental external impulse has been measured in a limited number of studies.  In the past, the accuracy of pressure transients measured in the thin spinal CSF layer was limited by the pressure transducers available; the fluid-filled catheters used were likely to dampen and lengthen the dynamic fluid impulse signal.  Contemporary miniature transducers offer the potential for indwelling measurements in large animals during injury.  Miniature and flexible transducers can be implanted through soft tissues and into thin fluid-filled spaces, with minimal alteration of the biomechanics of the system.  1.8 Research objectives and rationale The general objective of this research is to develop an understanding of how the CSF influences the extent of damage induced during the instant of a SCI, and how it responds to both the initial mechanical insult and the physiological processes that occur in the hours after such an injury.  The results of these studies will improve understanding of the biomechanics of SCI, and provide data which will enable SCI researchers to better understand the limitations of current models and to develop more biofidelic physical, animal and computational models.  Furthermore, it will provide an insight into several aspects of the injury sequelae that cannot be studied in the clinical population and that may have implications for the clinical and surgical management of SCI patients.  The objectives will be addressed in three phases comprising a total of five studies. Phase 1:  Characterising CSF pressure transients during the spinal cord injury event Objectives of Study 1:  Develop a large animal model, and associated instrumentation, of SCI capable of delivering an injury with human-like mechanical parameters to a system of human-like scale.  Specify a CSF pressure measurement and data acquisition system useful for both high-speed injury measurements, and measurements over extended periods. Rationale:  Biomechanical phenomena – and arguably biological phenomena as well – are best studied in models that are similar in scale to their human counterpart.  The need for a contemporary large animal 85 model of SCI to be used in pre-clinical testing is generally agreed upon in the SCI research community. Previous large animal injury devices have not been able to measure load or displacement.  Only two groups have measured CSF pressure transients during SCI, and the data reported were limited.  In Chapter 2, the first version of a new injury model is presented along with preliminary pressure results. Objectives of Study 2:  Characterise the transient CSF pressure distribution associated with injuries created with human-like mechanical injury parameters, and compare these to previously reported pressure tolerances of neural tissue. Rationale:  There is experimental and clinical evidence that central nervous system tissue is susceptible to damage by fluid pressure impulses.  At present, no studies have measured the transient CSF pressures created during a human-like SCI event.  The pressure transients associated with two injury severities created with human-like mechanical parameters, and a comparison to known pressure tolerances of neural tissue, are described in Chapter 3.  This chapter also presents the final version of the injury model. Phase 2:  Exploring the effect of CSF thickness and impact velocity on CSF pressure transients during the injury event Objectives of Study 3:  Design and construct a synthetic model which is anatomically and mechanically similar to the spinal cord – CSF – dura system, and investigate the influence of CSF thickness and impact velocity on transient pressure distribution, cord compression, and load transmission. Rationale:  While in vivo animal models are an invaluable model of the human condition, they offer limited opportunity to finely control independent variables and thereby study their effect on the dependent variables.  By employing a synthetic model, the thickness of the CSF and the impact velocity could be adjusted, and the effect of these adjustments on the pressure transients, cord compression and load transmission could be measured.  The results of this bench-top model are presented in Chapter 4. Phase 3:  Post-injury CSFP differentials before and after decompression, and the effect of decompression on spinal cord and thecal sac dimensions Objective of Study 4:  Profile the CSF pressure distribution in the 14 hours after injury, under conditions of sustained compression (eight hours) and decompression (six hours). Rationale:  The immediate post-injury phase is recognised as a critical phase in instituting neuroprotective strategies, yet little is known about the physical response of the cord and CSF in this time period.  CSF drainage has been proposed as a clinical management strategy to increase spinal cord perfusion.  This treatment would involve draining CSF from the lumbar region when some lumbar CSF pressure threshold is exceeded.  However, pressure differentials may exist across the injury site if there is residual cord compression and subarachnoid occlusion.  It is not possible to study these effects in the patient population 86 because (1) introducing catheters cranial to the injury creates risk of higher injury, and ventricular shunting is not indicated unless there is concomitant brain injury, and (2) the time lapse between injury, hospitalisation and stabilisation precludes study in the hyperacute phase.  Chapter 5 presents the profile of the CSF pressure cranial and caudal to an experimental SCI in a large animal over eight hours of residual compression followed by six hours of decompression. Objective of Study 5:  Profile the patency of the subarachnoid space (with respect to spinal cord deformation and swelling) in the hours after decompression. Rationale:  Early decompression is currently a treatment option for SCI patients with residual canal stenosis and spinal cord compression.  However, little is known about the immediate effects of decompression on the spinal cord morphology or the patency of the subarachnoid space.  In many patients, post-surgical magnetic resonance images show that the spinal cord has swollen and the thecal sac appears to be occluded within 24 to 48 hours after decompression.  It is not possible carry out serial imaging studies on patients during this time period due to their significant care requirements.  Spinal cord and dura morphology was evaluated, using ultrasound imaging, at the injury site for six hours after decompression of an experimental SCI in a large animal.  This study is presented in Chapter 6.  87 Chapter 2 A Large Animal Model of SCI to Measure CSF Pressure1 2.1 Introduction Spinal cord injury (SCI) often results in considerable permanent neurological impairment.  The handful of promising experimental treatments that have been tested in human clinical trials have failed to demonstrate convincing efficacy, prompting numerous researchers to identify areas where the process of translation from pre-clinical animal models to clinical application could be improved [491-494].  One issue of concern is the potentially important distinctions between the human spinal cord and the rodent or mouse spinal cord with which the vast majority of SCI research is conducted.  Not only may there exist unique anatomic and physiologic characteristics, but the sheer size and scale differences between rodent and human may be important.  By providing a more human-like anatomic scale and physiology, a large animal model may help to bridge the gap between basic science and clinical practice [374]. One potentially important shortfall of the rat and mouse species that are currently in prominent use is the lack of a cerebrospinal fluid (CSF) filled intrathecal space commensurate in size to its spinal cord [495], as well as the small size of the central nervous system (CNS) as a whole [395].  The CSF is commonly thought to protect the spinal cord during activities of daily living. However, several researchers have posited that tissue injury may occur remote from the site of mechanical insult as the result of a pressure or stress wave transmitted in the CSF and spinal cord tissue during SCI [1,3,162].  A pressure transient mediated injury mechanism could help to explain observations such as post-traumatic ascending myelopathies that can occur via an unknown etiology [170,173,174] and diffuse axonal injury remote to the primary mechanical insult observed in SCI patients [163] and in the spinal cord of infants with shaken baby syndrome [156,157].  Determining if the CSF has the potential to transmit injurious loads away from the primary injury site, and if so, under what conditions this can occur, could improve the selection of SCI models, increase understanding of primary and secondary cellular response, and allow interventions to be tested in a more clinically applicable experimental setting. The established rodent models of SCI are inappropriate for the study of CSF pressures associated with injury because the intrathecal space is extremely small and therefore transducers can only be placed in the cisterna magna and lumbar cistern [104,105,496].  Further, even transducers less than 1 mm in diameter would constitute a large fraction of the rodent CSF system total volume and are therefore likely to lead to changes to the mechanical behaviour of the cord, dura and fluid during impact.  In most cases  1  A version of Chapter 2 has been submitted for publication.  Jones CF, Lee JHT, Kwon BK and Cripton PA. Development of a large animal model to measure dynamic cerebrospinal fluid pressure during spinal cord injury. 88 the medium-to-large species that were previously used for weight-drop SCI models, such as cats [e.g. 277,366,367,369], dogs [e.g. 362,363,364], sheep [e.g. 218,370], and small non-human primates [e.g. 294,295,358,359], had spinal cord dimensions considerably smaller than humans.  Contemporary models using these species are largely limited to hemisection [e.g. 497,498] rather than contusion injuries, but contusion injuries are more clinically relevant as they represent the majority of human non-penetrating injuries [355]. Pigs have been used extensively for the study of  TBI and have a number of characteristics that make them useful for studying SCI, including a more human-like diameter of both the cord and intrathecal space [376].  Pigs have also been frequently used for models of SCI induced by ischaemia [377-379,381].  There are several reports of contusion [382,383] and clip compression [384,389] SCI models in pigs, although most do not provide a biomechanical analysis of the impact (i.e. impactor shape, energy, velocity, force).  The pig has spinal cord [384,398]  and CSF dimensions that are similar to an adult human [391-394] and is therefore suitable for characterisation of various aspects of the biomechanics of SCI, and in particular of the role of CSF during and immediately after the injury. Despite both historical and contemporary recognition of the potential for a mechanical role of CSF in SCI [1,3,280,281], there have been very few attempts to determine intrathecal fluid pressures either during experimental SCI in animals, or by using computational or physical models.  To the authors’ knowledge, only two groups have measured spinal fluid pressure in an animal model during experimental SCI.  Peak pressures of 50 mmHg and 150 mmHg were measured close to the injury site in single cats subjected to 15g-25cm and 20g-15cm weight-drop injuries [1,2].  In dogs subjected to closed-spine lateral impacts, peak CSF pressure ranged from positive peak 1960 mmHg to a negative peak of -630 mmHg, at 6 cm from the injury site for a single animal.  Both positive and negative pressure magnitudes decreased with distance from the impact site, reaching 27.5 mmHg and -19 mmHg at 44 cm from the impact [3]. Spinal CSF pressures have also been measured with limited success in cadavers [289] and pigs [288,487] during simulated whiplash events using rigid needle-mounted transducers.  Computational (finite element) models of SCI typically have not included fluid elements [e.g. 461,495,499], due in part to a lack of data for validation. Recent availability of sub-1 mm diameter, flexible, low-range pressure transducers with high frequency response, based on fibre optic and strain gauge technology, has made transducer placement in the spinal intrathecal space a realistic, although still challenging, prospect.  Such transducers have been applied to intracranial fluid pressure measurement in various TBI models such as fluid percussion, cortical impact, contusion and acceleration and blast wave models [290,314,337].  However, they have not been used in the spinal intrathecal space or for studying SCI.  This may be partly attributed to the lack of an appropriate contemporary large animal SCI model with proportionately large intrathecal sac 89 dimension.  The objectives of this study were to establish an in vivo porcine model of SCI suitable for studying the mechanics of CSF pressure during injury, and to determine changes in pressure magnitude at several locations cranial and caudal to the injury site during the SCI event using implanted miniature intrathecal pressure transducers. 2.2 Methods The experimental protocol was approved by the Animal Care Committee of the University of British Columbia and complied with the guidelines and policies of the Canadian Council on Animal Care. 2.2.1 Animals Seventeen female Yucatan miniature pigs (Memorial University of Newfoundland, Canada; Sinclair Bio-Resources, Windham, ME, USA) were group housed and acclimatised at our facility for at least one week before surgery.  At surgery the animals’ mean age was 20.9 (SD 2.2) weeks, and mean weight was 22.5 (SD 2.0) kg.  Anaesthesia was induced with intramuscular (IM) injection of Telazol 4-6 mg/kg; animals were endotracheally intubated and maintained on isoflurane (2-2.5% in O2 or as required) during the procedure.  Mechanical ventilation was maintained at 10-12 breaths/min with a tidal volume of 10-12 mL/kg.  Analgesics (hydromorphone 0.15 mg/kg IM and morphine 1mg/kg IM) were administered before the surgery and as needed during the procedure.  Antibiotic (cefazolin, 20 mg/kg) was given intravenously (IV) immediately before surgery and again at four hours.  Hydration was maintained with IV lactated Ringer’s solution.  Core temperature was monitored with a rectal temperature probe and maintained at 38.5-39.5 °C with a circulating-water heating pad.  A catheter was placed in the femoral artery to monitor arterial pressure.  A urinary catheter was placed either urethrally or suprapubically, and allowed to flow freely.  Anaesthetic level was monitored and adjusted to maintain heart rate, blood oxygen and blood pressure within normal physiological limits.  All animals were subsequently used for separate studies where the SCI applied in this study was a required step in the protocol. 2.2.2 Injury device A custom modified weight-drop device (Figure 2-1 and Figure 2-2) was designed to impart the SCI at T10 using either a 50 g (Group A, n=5; Group B, n=6) or 100 g (Group C, n=6) weight released from 50 cm.  Because Group A animals were used for model development, we did not collect complete velocity, force and CSF pressure data for all animals (see Table 2-1); in all other respects Group A was identical to Group B.  The weights used were higher than would be expected for a bone fragment associated with a burst fracture, but corresponded to the upper range previously used for large animal experimental SCI [e.g. 218,276,358,360,366,500] and were lower than the estimated wet weight of human thoracolumbar vertebra (250-550 g) [465].  The height was selected to produce impact velocities 90 within the range estimated for bone fragments during thoracolumbar vertebral burst fractures, 1-7.5 m/s (Section 1.6.4)  [132,462]. The injury device was attached to the spine unilaterally at T9 and T12 with pedicle screws (4.5 x 35 mm, CD Horizon, Medtronic, Minneapolis, MN).  A titanium rod (5.5 mm diameter, 70 mm length, CD Horizon, Medtronic, Minneapolis, MN) bridging these screws was then rigidly fixed to the base of an articulating arm with three ball joints (Model 660 (modified), L.S. Starett, Athol, MA, USA).  This allowed the aluminum guide cylinder (12.7 mm internal diameter, 60 mm length) to be held perpendicular to the spinal cord, with its lowest end approximately 2 mm from the dura surface.  The weight consisted of a 12.6 mm diameter stainless steel cylinder instrumented with a load cell (range ±222 N, nonlinearity and hysteresis ±0.5% of full scale, nonrepeatability ±0.1% of full scale, LLB215, Futek Advanced Sensor Technology, Irvine, CA, USA).  A 9.53 mm diameter spherical stainless steel impact tip was fixed to the load cell.  This diameter closely matched the maximum lateral diameter of the thoracic dura, as measured by in vivo magnetic resonance imaging in three animals (unpublished data).  The load cell cable was guided in a channel running the length of the guide cylinder.  The weight drop was triggered by removing a pin that bridged the centre of the guide cylinder and held the tip of the weight at 50 cm from the dural surface. A high speed camera (Phantom V9, Vision Research Inc., Wayne, NJ, USA) with a low distortion lens (AF Zoom-Nikkor, Nikon, Tokyo, Japan) was used to track quadrant markers (12.7 mm diameter) that were rigidly attached to the top of the weight and the side of the guide cylinder.  Images with a field of view of approximately 50×275 mm were obtained at 5000 frames per second and 240×1344 resolution, for 1.1 seconds.  The high speed images were used to determine the velocity of the falling weight.  Force data and high speed video were acquired for Groups B and C, but not for Group A. 91   Figure 2-1  Photo of the components of the modified weight-drop injury device.  Inset: close-up of the load cell installed within the weight base, above the spherical impactor tip.  92   Figure 2-2  Schematic of front view (left) and side view (right) of the weight-drop device installed over four vertebrae.  2.2.3 Pressure transducers CSF pressures were measured with miniature fibre optic pressure transducers (Preclin 420LP and Samba202 Control Unit, Samba Sensors, Sweden) with a sensor tip of 0.36 mm diameter, and a bare fibre diameter and length of 0.25 mm and 50 mm, respectively.  The sensors had a measurement range of -37.5 (vacuum) to 262.5 mmHg, and an accuracy of ±0.38 mmHg +2.5% (of reading) up to 188 mmHg and +4% (of reading) above 188 mmHg.  The frequency response of this transducer is not published, but it has been used previously to measure in vivo response to blast overpressures [341,342] and experimental TBI [290,314]. In this model development study the number of pressure transducers inserted in each animal increased as our technique improved and obtaining a seal between the transducer and the dura became reliable.  In Group A we placed one (n=3) or two (n=2) transducers at a nominal 100 mm from the injury site.  In Groups B and C we inserted two transducers in every animal (n=12) at a nominal 100 mm from 93 the injury site.  These transducer sites are referred to as “cranial-far” and “caudal-far”.  In two Group C animals we placed an additional two transducers at a nominal 20 mm from the injury site.  These transducer sites are referred to as “cranial-near” and “caudal-near” (Figure 2-3). 2.2.4 Surgical protocol The animal was anaesthetised and the dura-enclosed spinal cord was exposed via a laminectomy from approximately T7 to T14.  The laminectomy was widened at T10 to facilitate passage of the weight. Pedicle screws were inserted unilaterally into T9 and T12, and the articulating arm fixed in place.  The animal was tilted, head down, and the pressure sensors were inserted in the locations described previously.  The dorsal dura was pinched with forceps and gently lifted, a small hole was made in the raised portion with a 20 Ga needle tip, and the sensor advanced 50 mm into the intrathecal space along the dorsal aspect of the cord.  A small cone-shaped bone wax plug was moulded around the fibre to plug the hole in the dura, and this assembly was sealed with cyanoacrylate adhesive gel.  The distance from the injury epicentre to the pressure transducer tip was measured with callipers.  The cranial-far and caudal-far transducer faced away from the injury site, while the cranial-near and caudal-near faced towards the injury site (Figure 2-3).  The animal’s head was raised to bring the thoracic spinal cord into a horizontal position and the guide cylinder was attached and aligned vertically using spirit levels.  Immediately prior to the injury, the animal’s ventilation was stopped to cease respiration motion and the trigger pin was removed to induce the injury, after which ventilation was resumed.  The weight was removed from the spinal cord five minutes after impact.  For the subsequent studies, five animals remained under anaesthetic for a further 11 hours and were then euthanised with an overdose of sodium pentobarbital (120 mg/kg IV) without waking, and 12 animals were woken from anaesthetic immediately after injury and then enrolled in a separate functional recovery study for three months, after which time they were euthanised.  This heterogeneity did not influence the present study, which focused on the pressure behaviour during injury only. 94   Figure 2-3  Photo showing the surgical site, with 4 pressure transducers implanted intrathecally, a widened laminectomy at the injury site (T10), and pedicle screws in T9 and T12 (top); same photo with overlay indicating locations of pressure transducer tips relative to the injury site (shown as a circle) (bottom). The two near transducers were only implanted for two of the Group C animals.  2.2.5 Data acquisition, analysis and statistics Pressure and load data were acquired at a sampling rate of 50 kHz with custom LabVIEW programs (V8.6, National Instruments, Austin, TX, USA) and post-filtered with a two-way low-pass 4th- order Butterworth filter with a 5 kHz cut-off frequency using custom Matlab code (V2008a, The Mathworks, Matick, MA, USA).  High speed video was captured using the camera software (Phantom V9.0.640, Vision Research Inc., Wayne, NJ, USA).  Data were synchronised via the video trigger pulse. The markers in the high speed images were semi-automatically tracked using TEMA software (V3.2-024, Image Systems AB, Linköping, Sweden) using the quadrant marker setting.  The location of each marker in the first video frame was manually identified by the operator.  The software algorithm searched within a defined area on each subsequent image to locate each marker based on dark and light areas forming a quadrant pattern.  The centre of symmetry of each marker was then identified using Gaussian derivative filters.  The software algorithm is independent of rotation and the scale of the marker. 95 For each animal in Group A the peak positive and negative pressures were determined for each implanted transducer.  In Groups B and C the following parameters were determined for each animal: impact velocity, peak load, and peak positive and negative pressures for each transducer.  All pressures are reported relative to the pressure value immediately prior to the injury event because the average CSF pressure at any point along the length of the spine is dependent on the vertical height relative to the highest point in the fluid system (i.e. the hydrostatic pressure) [100] as well as the effect of pressure fluctuations from the respiratory and arterial pulse cycles.  The injury input and output parameters and the pressure transducer locations for each experimental group are detailed in Table 2-1. Table 2-1  Input injury parameters, numbers of successfully recorded model assessment (output) parameters, and pressure transducer sites by experimental group.  Group A (n=5) Group B (n=6) Group C (n=6) Input parameters Drop weight (g) 50 50 100 Drop height (cm) 50 50 50 Model assessment measures Velocity - n=6 n=5 Displacement - n=6 n=5 Force - n=5 n=4 Pressure assessment measures Peak positive and negative pressure Cranial-far (n=4) Caudal-far (n=3) Cranial-far (n=6) Caudal-far (n=6) Cranial-far (n=6) Caudal-far (n=6) Cranial-near (n=2) Caudal-near (n=2)  To assess the general characteristics of the model, non-parametric descriptive statistics (median, 25th and 75th percentile) were determined for each parameter for Group A, B and C separately. Differences in injury parameters (impact velocity, peak load and peak CSF pressures) between Group B and C were assessed with Mann-Whitney U Tests.  Group A was omitted from this analysis because only CSF pressure was measured in these animals.  For animals in Groups B and C, Wilcoxon matched pairs tests were used to assess differences in peak positive pressure, peak negative pressure and the distance of the sensor face from the injury epicentre, for the cranial and caudal sites.  A p-value less than 0.05 was considered statistically significant. 2.3 Results All animals remained stable past the experimental endpoint of this study.  Two animals in the recovery study died within three days of surgery, one from aspiration pneumonia and the other from an unknown cause.  Both were from Group C.  The CSF loss during transducer insertion was visually estimated to be around 1-3 mL per animal.  Areas of haemorrhage were observed on the dorsal aspect of 96 the cord in all animals when the weight was removed.  This haemorrhage typically extended approximately 1 cm both cranial and caudal of the epicentre.  The haemorrhage pattern was often discontinuous, with the central area corresponding to the contact area of the impactor tip devoid of visible blood.  We considered the injuries for all groups to be “severe” due to the haemorrhage observed immediately after injury, the limited hind-limb motor function in pigs that were monitored for three months post-injury, and subsequent histology which showed little white matter sparing at the injury site [501]. The impact parameters and peak negative and positive pressures at the “far” location for each test are shown in Table 2-2.  As noted in Table 2-1 above, due to technical difficulties load data were not collected for one test in Group B and two in Group C, and displacement data were not available for one in Group B.  The pressure data for the “near” location are provided in Figure 2-6. 97 Table 2-2  Injury parameters and peak positive and negative CSF pressures (relative)  at the “far” location for instrumented SCIs with 50 g and 100 g weight-drop. Animal ID Weight Pre- impact Velocity Peak Load Distance cranial-far Distance caudal-far Peak –ve cranial-far Peak +ve cranial-far Peak –ve caudal-far Peak +ve caudal-far  (kg) (m/s) (N) (mm) (mm) (mmHg) (mmHg) (mmHg) (mmHg) Group A (50 g weight) P406 21.0 - - nom. 100 nom. 100 - - -2.4 95.3 P403 20.5 - - nom. 100 nom. 100 -2.7 57.3 - - P412 22.0 - - nom. 100 nom. 100 -2.5 201.4 - - P408 25.5 - - nom. 100 nom. 100 -9.6 34.3 -16.3 61.2 P1202 24.0 - - nom. 100 nom. 100 -52.6* 86.7 -22.2 131.1 Group B (50 g weight) P1219 23.5 2.25 50.2 112 -117 -3.2 80.7 -4.9 55.4 P1241 24.0 2.63 58.0 111 -99 -4.1 134.6 -3.2 201.9 P1216 23.5 2.72 82.1 112 -105 -17.9 115.4 -18.7 72.3 P1205 24.0 2.76 54.0 109 -110 -36.0 136.0 -9.6 81.8 P407 22.5 2.58 46.8 98 -118 -4.1 105.5 -5.5 130.4 P1080 25.5 2.26 - 138 -150 -5.0 40.5 -6.1 24.2 Valid N 6 6 5 6 6 6 6 6 6 Median 23.8 2.6 54.0 112 -114 -4.6 110.5 -5.8 77.1 25th%ile 23.5 2.3 50.2 109 -118 -17.9 80.7 -9.6 55.4 75th%ile 24.0 2.7 58.0 112 -105 -4.1 134.6 -4.9 130.4 Group C (100 g weight) P1242 20.5 2.70 139.5 94 -104 -51.9* 11.0 -28.6 18.4 P1262 17.5 2.77 165.2 83 -94 -27.5 71.2 -21.3 101.5 P1264 22.0 2.52 125.2 98 -100 -7.5 272.9* -28.0 208.5 P410 21.5 2.73 101.1 65 -88 -34.7 20.7 -31.2 48.2 P416 23.5 2.78 - 100 -100 -27.7 105.6 -17.6 69.4 P411 22.0 - - 91 -112 -9.4 187.2 -26.4 65.0 Valid N 6 5 4 6 6 6 6 6 6 Median 21.8 2.7 132.4 92 -100 -27.6 88.4 -27.2 67.2 25th%ile 20.5 2.7 113.1 83 -104 -34.7 20.7 -28.6 48.2 75th%ile 22.0 2.8 152.4 98 -94 -9.4 187.2 -21.3 101.5 *these values were outside the manufacturer’s published range for the transducer (-37.5 to 262.5 mmHg), but within the range of the data acquisition card.  Linearity in this range was verified against a reference transducer in the laboratory.   98 2.3.1 Contusion injury characteristics Group B had a wider range of impact velocity than Group C, but the nominal 2.6 m/s impact velocity was not significantly different between them (p=0.177) (Table 2-2).  The peak load for Group B ranged 46.8-82.0 N, and for Group C ranged 101.1-165.2 N.  The median maximum load was 145% greater for the 100 g Group C injuries than for the 50 g Group B injuries (p=0.016) (Table 2-2). Although the device did not have a mechanism to eliminate weight bounce, the impactor tip did not rebound above the original height of the spinal cord in any of the tests.  In some tests the impact tip resided lateral to the spinal cord after the impact (i.e. not on the dorsal surface of the dura), despite all attempts being made to align the impact centrally.  In these cases the spinal cord was apparently deflected laterally due to the rounded shape of the impactor tip. 2.3.2 CSF pressure The distance between the injury site and the transducer tip was significantly different between Groups B and C for the cranial-far transducer (p=0.004), but not the caudal-far transducer (p=0.065). In most animals the fluid pressure profile measured at the “far” location at the time of injury consisted of a small negative pressure pulse followed immediately by a positive pressure pulse (Figure 2-4 and Figure 2-5).  The magnitudes of peak negative and positive pressures had considerable variation between animals within each injury group (Table 2-2).  The magnitude of median peak negative pressure was greater for Group C (100 g) than Group B (50 g) for both cranial-far (6 times greater) and caudal-far transducers (4.7 times greater); however, the difference was significant for the caudal location only (cranial-far p=0.093, caudal-far p=0.004).  The median peak positive pressures were not significantly different for Group B and Group C (cranial-far p=0.818, caudal-far p=0.699). 99   Figure 2-4  CSF pressure: cranial-far (top) and caudal-far (middle), and load (bottom) for the Group B (50 g injury) animals.  Zero time is arbitrarily aligned at Caudal-far = ±3mmHg.  There were no significant differences between the peak pressures (positive and negative) measured at Cranial-far and Caudal-far locations.  The early rise of the impact force, relative to the CSF pressure, for P1219 may indicate contact between the impactor and the sides of the laminectomy prior to dura contact.  100  Figure 2-5  CSF pressure: cranial-far (top) and caudal-far (middle), and load (bottom) for the Group C (100 g injury) animals.  Zero time is arbitrarily aligned at caudal-far = ±3 mmHg.  There were no significant differences between the peak pressures (positive and negative) measured at Cranial-far and Caudal-far locations.  There was considerable variability in the pressure profiles; for example P1242 showed very little pressure change while P1264 exhibited a large fluctuation, yet the impact force profiles are fairly similar.  In the two animals in which additional pressure transducers were placed approximately 20 mm from the injury epicentre, the peak positive pressure recorded at the cranial- and caudal-near transducer locations exceeded the upper range of the pressure transducers and data acquisition system.  This was approximately 300 mmHg above the baseline CSF pressure.  Negative pressure pulses were not observed at either the cranial or caudal locations at the near measurement sites (Figure 2-6).  101  Figure 2-6  Plots of CSF pressures for the two tests from Group C in which two “near” transducers were implanted  (top: P1242; bottom: P1262).  The pressures at the “near” location exceeded the upper range of the pressure transducers.  Across groups B and C, on average the far-caudal transducer was placed slightly further away from the injury epicentre than the far-cranial transducer (p=0.002).  However, there was no significant difference between the peak pressures recorded caudal versus cranial to the injury site for experimental groups B and C (negative pressure p=0.583, positive pressure p=0.308). (Table 2-2) 2.4 Discussion A large animal model of traumatic SCI has the potential to more closely represent the scale and physiology of the human spinal cord, and therefore may be an important step in elucidating primary and secondary injury mechanisms, and evaluating the efficacy of treatments.  In this study the human-like size of the porcine thoracic spinal cord and intrathecal space (Section 7.3) was exploited to measure CSF pressures associated with spinal cord injuries from a novel weight drop device, by inserting miniature pressure sensors into the intrathecal space adjacent to the injury site.  The results suggest the miniature Yucatan pig is an appropriate model for studying CSF, spinal cord and dura interactions during injury and that with further development it may be an appropriate in vivo large animal model of SCI for many basic science applications where human-like dimensions and physiology are important. 102 2.4.1 Injury model Although previous SCI research has used a variety of medium-to-large species, contemporary work predominantly uses well established rodent models that are less ethically contentious and have less cost and complexity associated with surgery and care.  Pigs are well established as paediatric and adult models in TBI research [376] and as models of chronic spinal cord pathology [502-504].  They have also been used for ischaemic models of SCI [e.g. 379].  Isolated reports of mechanical SCI models in the pig include weight-drop [383], pneumatic impactor [382] and clip compression [384,389] techniques; however, to our knowledge, these models have not been characterised mechanically.  In our opinion, pigs are a suitable SCI model because their CNS is similar to humans in several respects including blood supply and flow characteristics [388,505], white and grey matter distribution, brain and spinal cord growth and development [506] and spinal skeletal similarities [386,464,507].  Specifically, the miniature Yucatan is an ideal breed for both acute and chronic studies as it is bred disease-free, and has a slow growth rate and low weight at full maturity [375].  However, there are also some notable deviations from human spine anatomy including a greater number of vertebrae [375] and, according to our observation and estimation, a greater amount of epidural fat and larger epidural vessels than in many humans. There are several challenges associated with the required surgery – it is technically demanding and lengthy, requiring trained surgical staff and close veterinary monitoring throughout.  Care must be taken to avoid the large epidural veins which are prominent in the anterolateral canal, and to avoid breach of the canal wall when placing the pedicle screws.  There is potential for considerable CSF loss during transducer placement; care must be taken to minimise the size of the puncture hole, and to introduce the transducer and apply sealant rapidly.  The bone wax “plug” was essential to form a patent seal and to prevent damage to the transducer upon its removal.  Larger animals are also more susceptible to infection than rodents, a consideration that is particularly important for protocols with long-duration anaesthetic and for survival studies.  All components of the injury device that reside in the surgical field can be autoclaved or soaked in cold sterile solution. Mounting the device to the spine served two purposes.  Firstly, it located the weight at a fixed distance from the dural surface throughout the large dorsal excursion associated with respiration in the ventilated animal.  Hung et al. [369] induced a unilateral pneumothorax to reduce this motion in cats, but the methodology used to combat this motion is otherwise rarely described for apparatus that were fixed to the operating table.  This feature also makes the injury device well suited to experiments in which prolonged compression of a consistent magnitude is desired after the dynamic injury.  Secondly, the pedicle screw and rod construct eliminated relative motion between the fixed vertebra and likely improved consistency of the energy delivered to the spinal cord; others have placed a saddle under the spinal cord [296] or rigid supports under the transverse processes [396]. 103 The impactor tip had approximately the same diameter as the maximum lateral diameter of the cord; it is not known how this compares to other devices because cord dimensions were not commonly reported.  The spherical tip ensured maximum tip advancement without impingement on the lateral canal walls; however, the relative instability of the contact between the convex impactor tip and dural surface may have caused lateral displacement of the spinal cord relative to the tip.  Although great care was taken to centre the guide cylinder relative to the lateral borders of the dura, this instability would have been exaggerated by any lateral misalignment of the guide cylinder.  While the final resting position might not be indicative of alignment during the impact event, the lateral motion may have influenced the resultant injury in some animals.  Previously reported weight-drop devices commonly used a secondary “impounder” resting on the cord [e.g. 294] presumably to aid alignment of the falling weight.  However, this reduces the energy transferred from the falling weight to the spinal cord [364] and even light weight static impounders alter the size and shape of the intrathecal space and spinal cord prior to injury [369], which was undesirable in the current study. It is difficult to compare the impact velocities and peak loads obtained in this study with previous published results because very few large animal studies report velocity, peak impact force and spinal cord dimensions.  In our study the load cell provides valuable information regarding the relative force profiles applied for the two injury severities; the significant difference between peak impact forces for Groups B (50 g) and C (100 g) was expected due to the larger mechanical energy associated with the larger mass. Due to space and weight restrictions, the impactor did not incorporate an accelerometer for inertial compensation; however, the mass of the impactor tip (below the load cell) was 3.45 g and therefore from Newton’s second law this would require a correction of less than 8%.  Variation of peak force within the experimental groups can be attributed to factors such as differing tissue mechanical responses due to variation in the size of the spinal cord, dura and ligamentous tissues, presence of different amounts of CSF, and any deviations of the impactor tip from the centerline. We did not report a measure of dural displacement in this study because we thought that any lateral deviations of the impactor tip would confound the measurement. 2.4.2 CSF pressures While the CSF likely provides mechanical protection for the central nervous system during minor whole body perturbations, a mechanical insult which imparts a large energy to the system may produce a pressure wave in the CSF and in the fluid and solid constituents of the spinal cord that transfers energy (and deformation) away from the site of the injury, thus disrupting cellular function at locations remote from the injury.  Despite early recognition of this potential [1,3] further focus in published reports is limited to very few descriptions of diffuse axonal injury (DAI) remote to the site of mechanical insult. DAI has been observed in the cervical spinal cord following experimental TBI in ferrets subjected to 104 cortical impacts [508] and in rabbits subjected to rotational head acceleration [509].  Human occurrences are limited to anecdotal evidence of DAI in the cervical cord associated with shaken baby syndrome [156,157], and evidence of axonal injury at sites remote from the focal injury in SCI patients who died between 4 hours and 6 weeks after injury has been reported [163].  Recently, Czeiter et al. [162] observed that impact induced TBI was capable of evoking traumatic axonal injury in the cervical and thoracolumbar spinal cord of rats.  Observing that the majority of the injured cells in the thoracolumbar region were close to the surface of the cord, they proposed that in addition to the commonly cited axonal stretch mechanism, a shock wave travelling through the CSF could have contributed to this cellular injury. The peak CSF pressures measured in the current study were several orders of magnitude higher than normal resting pressures.  Baseline CSF pressures have been measured in the rat cisterna magna 6-8 mmHg [496] and lumbar intrathecal sac 0.7-7.4 mmHg [105], 4.18 mmHg [104]; in cats, 8.7 mmHg [101] and in dogs: cervical 8.8 mmHg and lumbar 10.0 mmHg [100].  Transient pressure increases in humans are associated with the Valsalva manoeuvre, 10.5 mmHg [64] and coughing, 40 mmHg [63].  To the authors’ knowledge, there is no established pressure tolerance value for the bulk spinal cord or its cellular constituents.  However, animals exposed to external blast overpressures ranging from 75 to 7500 mmHg have exhibited reduced performance in physical tests and degeneration of cerebral cortex neurons [340], injury to neuronal and glial cells, and brain edema [344-349], brain haemorrhage and edema [342], morphological changes to neurons [351], damage to central visual pathways [343] and elevated intracranial pressure and impaired cognitive function [341].  While neural tissue in different locations may have different injury susceptibility due to differences in microvasculature, metabolic processes and cellular distribution, these studies suggest that the peak pressures observed in the current study may approach or exceed pressure tolerances and thus may contribute to the overall injury severity sustained by the animals. To our knowledge there have been only two single measurements of spinal CSF pressure during an open experimental SCI.  Hung et al. [1] reported a peak pressure of 50 mmHg approximately 25 mm cranial to the injury site in a cat weight-drop model using a 15g-25cm injury, with a 5 mm diameter cylindrical weight.  The same group, using a similar weight-drop cat model, reported a 150 mmHg peak pressure at 15 mm caudal and cranial to the injury site, in response to a 20g-15cm injury [2].  Differences in animal size, weight-drop parameters and pressure transducer placement make direct comparison to our data difficult; however, it is noted that the peak CSF pressures measured at 100 mm from the epicentre in our experiments were of similar or greater magnitude to those measured previously.  The peak pressures in the current study exhibited wide variation which may be due to variation in the peak impact load, baseline CSF pressure, CSF layer thickness and the nature of the contact between the spherical impactor tip and the dura.  In the two animals with transducers 20 mm from the epicentre, the “out-of-range” 105 readings imply that the pressures at the equivalent distance were at least an order of magnitude higher than those measured in the cats.  This may be due to the higher injury energy used in the current study, and also the differing size of the cat and pig anatomy.  Further, in those studies the CSF pressure was measured via an implanted “fluid filled needle” and tube attached to an external pressure transducer.  This liquid measurement path probably caused attenuation of the pressure signal which did not occur in the current study because the sensing face was in direct contact with the CSF.  The pressure trace provided in Hung et al. [1] indicates that the positive pressure pulse was followed by a smaller negative pressure pulse of around -16 mmHg.  It is not known why negative pressures occurred after the initial positive peak in Hung’s study, while in our study the negative trough preceded the positive peak. The most comprehensive published study of  spinal CSF pressure transients used a closed-spine lateral weight-drop SCI model in dogs [3].  Pressures ranging from -630 mmHg to 1960 mmHg were measured at 60 mm from the impact site, with diminishing ranges of approximately -400 to 288 mmHg at 150 mm, and -28 to 19.0 mmHg at 440 mm from the impact site.  These pressures are considerably higher than those of the present study, particularly at the measurement locations closest to the impact.  The closed-spine model imparts the SCI by dropping a large mass on the intact spinal column and so it is difficult to compare the injury severity to that of the current study or to human SCI.  Wennerstrand et al. [3] also used a fluid-filled catheter attached to an external transducer and accounted for the damping by applying a correction factor determined from a synthetic model of the system; however, it is unclear how well this replicated the animal tests and therefore if the correction factor produced accurate estimates of the CSF pressure within the thecal sac. Other studies that have measured spinal pressure associated with injury have either used physical models without CSF, or have not replicated the kinetics and/or kinematics that produce human SCI. Yoganandan and colleagues [469] placed a pressure transducer instrumented gelatine spinal cord in cadaver cervical spines with a C4 laminectomy and subjected them to a number of weight-drop tests. They recorded pressures between 240-400 kPa (1800-3000 mmHg) at the injury epicentre, and 30-50 kPa (225-375 mmHg) at one vertebral level rostral/caudal, for 200-600 g-cm impacts.  The authors did not provide separate weight and height measures for the weight-drop, but the pressure values reported are of similar magnitude to the current study.  Other researchers have measured cervical CSF pressures in whiplash models using pigs [287,288] and post mortem human subjects [289].  The former recorded pressures from -100 to 150 mmHg and observed that higher pressures were associated with the location of neuron membrane leakage; these values are within the range measured in the current study.  The latter study reported similar pressure amplitudes (0 and 220 mmHg); however, the vascular and CSF systems were not pressurised in these post mortem human subject tests. 106 In characterising biomechanical systems it is desirable to alter the natural state as little as possible. In this experiment, a long narrow laminectomy and subsequent removal of ligamentum flavum and epidural fat were required.  This, along with the addition of cyanoacrylate gel to the dural surface, alters the dorsal boundaries of the system and may influence the dynamic response of the dura and CSF. However, the open surgical approach allowed the introduction of miniature pressure sensors which permitted measurement of pressure in direct contact with the fluid.  The transducers’ size, weight and flexibility minimised possible effects of transducers on the mechanical response of the system.  It is possible that soft tissue came into contact with the pressure sensor face during the injury; however, the sensing face on the cross-section of the fibre was some distance from the injury site where large tissue deformations are less likely, and its small size and flexibility probably allowed it to bend in a similar manner as the soft tissue.  The effect of the anaesthetic protocol on the process of CSF formation and resorption is unknown but it is unlikely that this affected the pressure transients.  Finally, the weights used to impart the injury were greater than would be expected of a bone fragment retropulsed into the canal during a burst fracture SCI, and reducing this weight would enable more human-like injury parameters to be used. 2.5 Conclusion  A large animal experimental SCI model and injury device were developed and successfully used to directly measure intrathecal CSF pressure at the instant of the injury event using miniature flexible pressure transducers.  The peak magnitudes measured at approximately 100 mm from the injury epicentre were at least one order of magnitude above baseline.  Pressures measured at 20 mm from the injury epicentre exceeded the range of the pressure transducer.  Further studies utilising refined weight-drop parameters to better approximate a human-like SCI and pressure transducers with a higher range are needed to further evaluate the contribution of dynamic CSF pressure to primary injury.  107 Chapter 3 CSF Pressure during Contusion-type SCI2 3.1 Introduction Despite concentrated research efforts over the past three decades, spinal cord injury (SCI) continues to be a devastating and permanent condition.  While a number of treatments that showed promise in pre-clinical studies have been tested in human clinical trials, none has had demonstrable efficacy [374].  This lack of effective translation of treatments from laboratory models to bedside has been attributed, in part, to important differences between the human injury and the rodent models commonly used to represent it.  While the rat contusion model replicates some features of human SCI [510], there are substantive differences in neuroanatomy and behavioural outcome [511].  One clear disadvantage of the rat model is the large discrepancy in scale relative to human anatomy [512]. Recognising this, there is a strong sentiment within the SCI research community that establishing pre- clinical treatment efficacy in a suitable large animal model would be an advantageous step in the appropriate pre-clinical evaluation of treatments [374].  This size discrepancy has implications for the selection of suitable injury parameters and the subsequent mechanical and biological response of the system to the injury and to treatment agents.  We propose that an important and under-emphasised aspect of this scaling issue is the lack of a cerebrospinal fluid (CSF) layer of the same relative dimension as in humans. Humans have a CSF layer surrounding the spinal cord that is on the order of 2-4 mm thick in the thoracic region [391,394].  Previous ex vivo and in vivo studies have noted the potential for this fluid layer to be both mechanically protective [e.g. 1,280,281] and potentially injurious [1,162,285] depending on the nature and magnitude of the mechanical loading.  It has been proposed that a pressure wave of sufficient magnitude travelling away from the mechanical impact may injure cells some distance from the impact [1,162].  Axonal injury remote from the site of mechanical impact has been observed clinically [157,163], and in vivo and in vitro neural tissues are adversely affected by experimental fluid impulses [e.g. 313,332,342].  Further, some cases of post-traumatic ascending myelopathy [170,173,174] and SCIs which occur without evidence of  bony or ligamentous abnormality [161] do not have a well defined origin and could be partly due to a pressure-induced primary injury mechanism occurring at, or some distance from, the main mechanical insult.  Despite this potential “over-pressure” mechanism, animal, cadaver, and computational models that are used to study the mechanisms of SCI and treatment commonly do not include a fluid layer, and only two groups have measured CSF pressure during an experimental SCI [1-3].  2  A version of Chapter 3 has been submitted for publication.  Jones CF, Lee JHT, Burstyn U, Okon, E, Kwon BK and Cripton PA. Cerebrospinal fluid pressures during dynamic contusion-type spinal cord injury in a pig model. 108 The study presented in Chapter 2 demonstrated that measuring CSF pressure in a miniature pig model of SCI is feasible and provided some evidence that fluid pressures occurring at the instant of SCI may be sufficient to contribute to neural injury.  However, that preliminary study had several limitations.  Firstly, due to the requirements of concurrent studies using the same animals, the weight and height combinations used to impart the injury were not based on biomechanical studies of burst fracture processes.  Secondly, only two animals had pressure transducers placed near to the injury site; furthermore, the pressures measured at this location exceeded the range of the transducers used.  The previous study also highlighted several ways in which the injury device could be refined to improve its robustness and also the repeatability of the impact velocity obtained.  Finally, the preliminary study did not include histological analysis of the spinal cord tissue viability.  Therefore, the objective of this study was to characterise the CSF pressure waves associated with experimental SCIs with weight-velocity combinations approximating moderate and high human-like severity, at several locations, and to determine if these are sufficient to contribute to neural tissue injury some distance from the primary mechanical insult. 3.2 Methods The experimental protocol was approved by the Animal Care Committee of the University of British Columbia and complied with the guidelines and policies of the Canadian Council on Animal Care. 3.2.1 Animals Fourteen female Yucatan miniature pigs (~20 kg, Sinclair Bio-Resources, Windham, ME, USA) were group housed and acclimatised at the facility for at least one week before surgery.  Animals were assigned to high severity injury (n=6), moderate severity injury (n=6) and sham (n=2) groups. Anaesthesia was induced with Telazol (4-6 mg/kg IM), xylazine (0.6 mg/kg) and atropine (0.02 mg/kg, IV), animals were intubated, and maintained on isoflurane (2-3.5% in O2) with mechanical ventilation (10-12 breaths/min, tidal volume 10-12 mL/kg).  Analgesics (hydromorphone 0.15 mg/kg IM or morphine 1 mg/kg IM) were administered before surgery and thereafter every 3-4 hours; antibiotics (cefazolin, 20 mg/kg IV) were administered before surgery and then every 4 hours.  Blood pressure was monitored using Doppler ultrasound and a cuff on the forelimb.  All animals received lactated Ringer’s solution (IV).  Temperature was monitored via a rectal probe and maintained at 37.5-38.5 °C with a circulating-water pad.  Catheters were placed in the left carotid artery and external jugular vein to monitor central arterial and venous pressure.  The urinary bladder was catheterised using an 8 French Foley catheter.  Isoflurane concentration and fluid rate were adjusted as needed to maintain systolic arterial blood pressure within normal physiologic parameters. 109 3.2.2 Injury device The SCI was imparted with a custom modified weight-drop device comprising a 20 g weight released from a height of 25 cm (moderate severity) or 125 cm (high severity), replicating the bone fragment weight and velocity associated with burst fractures [132,462].  The injury device (Figure 3-1) was attached to the spine unilaterally at T10-T13 with pedicle screws and a bridging titanium rod (screws: 4.0x26/28 mm; rod: 3.5 mm diameter/150 mm length; Vertex Reconstruction System, Medtronic, Minneapolis, MN, USA).  This construct was fixed to the base of an articulating arm (Model 660 (modified), L.S. Starett, Athol, MA, USA) which enabled the aluminum guide rail (N17, Igus, Concord, ON, Canada) to be placed orthogonal to the spinal cord, resting on the remaining lateral borders of the vertebrae.  Guide rails of length 0.6 m and 1.5 m were used.  The weight consisted of a rapid-prototyped hollow cylinder (material: ABS-M30, sealed with acrylic lacquer) with a linear bearing attachment (aluminum/plastic, Drylin N17, Igus, Concord, ON, Canada) and was instrumented with a load cell (range ±222.41 N, LLB215, Futek Advanced Sensor Technology, Irvine, CA, USA) to which a 3/8” diameter cylindrical impact tip (material: ABS-M30, sealed with acrylic lacquer) with a 45°/1 mm bevelled edge was fixed.  This diameter closely matched the lateral diameter of the thoracic dura, as measured by in vivo magnetic resonance imaging performed on three animals in a separate study (unpublished data).  The impact tip height was set using a custom measuring tool and the weight was released by a latching solenoid (STA151082-234-1002, cage 81840, Saia-Burgess, Vandalia, OH, USA) with associated custom electronics. A high speed camera (PhantomV9.1, Vision Research, Wayne, NJ, USA) was used to track quadrant markers rigidly attached to the weight and guide rail (field of view ~50×275 mm, 5500 frames per second, resolution 240×1344).  The images were used to determine the impact velocity and the dorsal dura displacement during injury.  110  Figure 3-1  Schematic of front view (left) and side view (right) of the weight-drop injury device installed on vertebrae T10-T13; the injury was centred on the T11 vertebral level.  3.2.3 Pressure transducers CSF pressure was measured at four locations with miniature fibre-optic pressure transducers (Preclin 420LP/360HP with Samba202 control unit, Samba Sensors, Sweden) (Table 3-1).  The frequency response of these transducers is not published by the manufacturer but they have been used previously to measure high rate in vivo and atmospheric pressure impulses during experimental blast and TBI [290,314,341,342].  The transducer consists of 50 mm of bare glass fibre, then 50 mm of Teflon coated fibre, and 6 metres of plastic coated cable.  Two high range pressure transducers were placed at 30 mm from the injury site (cranial-, caudal-near), facing towards the impact locations, and two low range transducers were placed at 100 mm from the injury site (cranial-, caudal-far), facing away from the impact location (Figure 3-2).  Low range transducers were also used to measure the arterial and venous pressures.  The Samba 202 control units were sampled at 40 kHz, from which an analog signal was transmitted to a data acquisition system. 111 Table 3-1  Pressure transducer specifications  (Samba Sensors, Gruvgatan 6, Sweden)  Preclin420LP (low range) Preclin360HP (high range) Range (mmHg) 37.5 – 262.5 -75 – 3500 Accuracy (mmHg) ±0.38 plus ±2.5% of reading ± 4% of reading Sensor tip diameter (mm) 0.42 0.36 Fibre diameter (mm) 0.25 0.40 Fibre length (mm) 50 50 Temperature coefficient (mmHg/ºC, 20-45ºC) <0.15 <1.95 Coating nil radiopaque coating   Figure 3-2  Photo (top) and overlay (bottom) indicating the location of the four intrathecal pressure transducers and pedicle screws; the injury was centred on the T11 vertebral level.  3.2.4 Experimental protocol The animal was anaesthetised and the spinal cord was exposed via a laminectomy from T4 to L4. The laminectomy was widened at T10-T11 to ensure that the impactor tip did not strike the bony edges during impact.  Pedicle screws were inserted and the articulating arm attached.  The dura was exposed and the surgical table was tilted so that the animal’s head was angled downwards to reduce hydrostatic pressure at the pressure transducer insertion point.  The desired entry point for the subarachnoid pressure transducers was measured with callipers, then the dorsal dura was gently lifted with forceps at this 112 location and a small hole made with a needle tip.  The transducer was rapidly introduced and advanced 50 mm into the intrathecal space (to the Teflon coated section), on the dorsal aspect of the cord.  The dural hole was sealed around a small bone wax plug moulded to the transducer at the junction of the bare fibre and Teflon coating, using cyanoacrylate adhesive gel.  After all four transducers had been placed, the animal’s head was raised to bring the thoracic spinal cord into a horizontal position.  The guide rail was attached and aligned vertically such that the weight tip was centred over the T11 vertebral level.  The animal’s ventilation was held to cease respiration motion and the solenoid was activated to impart the injury.  The ventilation was resumed within three seconds after injury.  As part of a separate protocol the animals remained under anaesthetic for 14 hours post-injury with 100 g compression for 8 hours and then 6 hours without compression.  Animals were euthanised with an IV injection of sodium pentobarbital. The two sham animals received all surgical procedures except for the injury. 3.2.5 Histology The spinal cord was harvested rapidly 14 hours after injury, and a 5 cm segment, with the dura intact and centered on the injury epicentre, was immersed in 4% phosphate buffered paraformaldehyde at 4 °C for 72 hr.  Segments were cryoprotected in three 0.1 M phosphate buffered sucrose solutions (24%, 18% and 12%) for 2–3 days each, or until the tissue sank to the bottom of the container.  The dura was dissected from the spinal cord and 1 cm segments were embedded in Tissue Tek OCT (Sakura Fintek Inc, Torrance, CA, USA).  Transverse sections 200 µm thick were cut on a cryostat (Microm HM505E, Waldorf, Germany) and every second section mounted on slides, then stored at -80°C.  Sections were stained with Eriochrome Cyanine and counter-stained with Neutral Red [513] to visualise the spared white and grey matter.  For the analysis of lesion size, every fourth section (i.e. 1600 µm apart) starting from the estimated lesion centre was photographed at 5× objective with a microscope (DM5000B, Leica, Wetzlar, Germany), digital camera (DFC420, Leica) and software (Leica Application Suite, V3.1.0). Images for each section were merged in Adobe Photoshop (CS5, San Jose, CA, USA), the lesion area and total spinal cord area were manually segmented in the digital images using Analyse software (V10.0, Analyse Direct, Overland Park, KS, USA), and the percent “spared” tissue [(total-lesion)/total)] was calculated for each section.  The “epicentre” was defined as the cross-section with the least amount of white and grey matter sparing (i.e. the greatest extent of parenchymal damage). 3.2.6 Data acquisition, analysis and statistics Pressure, load and camera synchronisation data were acquired with custom Labview (V8.6, National Instruments, Austin, TX, USA) programs at 50 kHz then post-filtered and processed with custom Matlab (V2008b, The Mathworks, Matick, MA, USA) programs with a two-way 4th-order Butterworth filter with 5 kHz low-pass cut-off frequency.  High speed video was captured with Phantom 113 (V9.0.640, Vision Research Inc., Wayne, NJ, USA) and distortion corrected with a camera calibration routine [514].  Markers were tracked using TEMA software (V3.2-024, Image Systems AB, Linköping, Sweden) using the quadrant marker setting (see Section 2.2.5). For each test we determined mechanical parameters including peak impact force, force impulse, maximum dorsal dural displacement, impact velocity; and pressure parameters including peak positive and negative pressure, impulse, wave speed and attenuation ratio.  All CSF pressures are reported relative to the pre-injury value to eliminate the effect of hydrostatic pressure variation [100] and respiratory and vascular pulsations.  The wave speed was defined as the delay between the pressure peaks at the “near” and “far” locations, divided by the 70 mm transducer separation.  The attenuation ratio was defined as the ratio of the “near” to “far