Biomechanics of the human intervertebral disc: a review of testing techniques and

Many experimental testing techniques have been adopted in order to provide an understanding of the biomechanics of the human intervertebral disc (IVD). The aim of this review article is to amalgamate results from these studies to provide readers with an overview of the studies conducted and their contribution to our current understanding of the biomechanics and function of the IVD. The overview is presented in a way that should prove useful to experimentalists and computational modellers. Mechanical properties of whole IVDs can be assessed conveniently by testing ‘ motion segments ’ comprising two vertebrae and the intervening IVD and ligaments. Neural arches should be removed if load-sharing between them and the disc is of no interest, and specimens containing more than two vertebrae are required to study ‘ adjacent level ’ e ﬀ ects. Mechanisms of injury (including endplate fracture and disc herniation) have been studied by applying complex loading at physiologically-relevant loading rates, whereas mechanical evaluations of surgical prostheses require slower application of standardised loading protocols. Results can be strongly in ﬂ uenced by the testing environment, preconditioning, loading rate, specimen age and degeneration, and spinal level. Component tissues of the disc (anulus ﬁ brosus, nucleus pulposus, and cartilage endplates) have been studied to determine their material properties, but only the anulus has been thoroughly evaluated. Animal discs can be used as a model of human discs where uniform non-degenerate specimens are required, although di ﬀ erences in scale, age, and anatomy can lead to problems in interpretation.


Introduction
Intervertebral discs (IVDs) are pads of fibrocartilage which lie between the vertebrae of the spine. They allow the vertebral column to bend and twist (Bogduk, 2005;Humzah and Soames, 1988), and distribute compressive loading on the adjacent vertebral bodies (Adams and Roughley, 2006). The mechanical properties of discs are important because human lumbar IVDs are often physically disrupted (Vernon- Roberts et al., 1997), which may give rise to degenerative changes (Adams and Dolan, 2012;Adams and Roughley, 2006;Ferguson and Steffen, 2003) and to chronic back pain (Cheung et al., 2009;de Schepper et al., 2010). Accurate mechanical characterisation of the IVD is required to analyse injury mechanisms, and to understand how ageing and degeneration can degrade the material properties of a disc's component tissues (Ferguson and Steffen, 2003;Shan et al., 2015;Vernon-Roberts and Pirie, 1977), increasing vulnerability to injury. Other major motivations to study disc mechanics are to develop and test surgical implants such as prosthetic intervertebral discs (Cunningham et al., 2003;Lee and Goel, 2004;Lemaire et al., 1997), and to obtain accurate material properties for input into finite element models of the spine (Schmidt et al., 2013). A focus has also been on high strain-rate injuries, for example vehicle accidents, airplane ejections, and blast-related events (Panzer et al., 2011;Yoganandan et al., 1989).
The purpose of this article is to present our current understanding of the mechanical behaviour of the IVD, evaluate the methods used to test the IVD and its component tissues, and summarise the results of such tests in a manner that is useful to other experimentalists and computational modellers. The focus of the review is adult human lumbar discs. The very large number of studies concerning the human IVD poses a challenge of scale when conducting a review of this nature, and so for reasons of brevity we chose not to compare human and animal discs. However, we believe that studying animal tissues, which are often uniform and non-degenerated, can help in understanding how aspects of testing techniques (such as speed of loading, or complex loading) influence the results of experiments on human tissue, and for this reason some data from animal studies are included here. Cadaveric discs remain the closest available representation of living human discs, and so results from cadaveric studies can most easily be translated into the clinical setting. Only meso-or macro-scale tests are considered, rather than nano-or molecular level tests.

Disc composition and functional anatomy
A motion segment consists of two vertebrae and an IVD (Fig. 1). The IVD has three main components (Fig. 2): a soft, deformable, nucleus pulposus (NP), which is surrounded by the fibrous concentric layers of the anulus fibrosus (AF), and bonded above and below to adjacent vertebral bodies (VBs) by the thin layers of the cartilaginous end plates (CEPs). Lumbar IVDs are the largest avascular organs in the human body. Each of the three component tissues will be considered in turn.

Nucleus pulposus (NP)
The NP is a gelatinous structure that accounts for 40-50% of the volume of the adult disc (Bayliss and Johnstone, 1992;Iatridis et al., 1996;Pooni et al., 1986) and 25-50% of the transverse cross-sectional area (Farfan et al., 1970;Nachemson, 1960;Perey, 1957). The NP has such a high water content that it exhibits a hydrostatic pressure which increases in response to compressive loading (Keyes and Compere, 1932;McNally and Adams, 1992), and this pressure generates tension in the surrounding AF (Nachemson, 1963). Its main constituents are proteoglycan, collagen and water (Antoniou et al., 1996). Proteoglycan accounts for 35-65% of the dry weight of the NP, and is important for binding water into the tissue (Dickson et al., 1967;Iatridis et al., 1996;McDevitt, 1988). Fine collagen type II fibrils, which account for 5-20% of dry weight, provide a loose, 3-dimensional fibre network, which holds the nucleus together (Eyre, 1988;Inoue and Takeda, 1975). The remainder of the dry weight of the NP is non-collagenous proteins and elastin (Eyre, 1988). NP water content decreases with age, from approximately 90-70% from the ages of 1-80 years old (Antoniou et al., 1996;Gower and Pedrini, 1969;Kraemer et al., 1985), reflecting a similar decrease in proteoglycans.
In a healthy lumbar disc, in vivo pressures in the nucleus are between 460 and 1330 kPa in the seated position, 500 and 870 kPa in the standing position, and 91 and 539 kPa when lying either prone or supine Morris, 1964, 1963;Sato et al., 1999;Wilke et al., 1999). The highest pressure in the nucleus (2300 kPa) was recorded in a standing subject who was flexing forwards holding a 20 kg mass (Wilke et al., 1999).

Anulus fibrosus (AF)
The AF is made up of 15-25 concentric layersthe lamellaewhich are approximately 0.05-0.5 mm thick, of increasing thickness from outer to inner (Cassidy et al., 1989;Inoue and Takeda, 1975;Marchand and Ahmed, 1990). Approximately 48% of the lamellar layers are circumferentially incomplete and the percentage of incomplete layers increases with age (Marchand and Ahmed, 1990). Each layer consists of coarse and strong collagen type I fibre bundles, as in tendon, with their orientation alternating between ± 25-45°in relation to the transverse plane (Fig. 2c). The angle of inclination increases towards the centre of the disc (Cassidy et al., 1989;Horton, 1958;Hsu and Setton, 1999;Marchand and Ahmed, 1990;Pooni et al., 1986). A trans-lamellar, collagen-based bridging network provides shear resistance between adjacent lamellae (Adam et al., 2015;Melrose et al., 2008;Pezowicz et al., 2006;Schollum et al., 2008;Yu et al., 2007Yu et al., , 2005. The complex arrangement of collagen fibres in the AF enables it to develop tensile, 'hoop'or circumferentialstress due to the pressure in the nucleus.
In the healthy IVD, the AF contains 65-70% water. Dry weight is approximately 20% proteoglycan, 50-70% collagen, and 2% elastin   (Adams et al., 1977;Antoniou et al., 1996;Buckwalter, 1995;Mikawa et al., 1986;Yu et al., 2005). Moving from outer to inner AF, proteoglycan, water, and collagen type II content increases, whereas collagen type I content decreases. Mechanically, type I collagen provides strength in tension, as in tendon. Type II collagen forms a fine meshwork that binds with proteoglycans -and hence water -thus enabling the tissue to withstand large compressive forces, as in hyaline cartilage (Adams et al., 1977;Eyre and Muir, 1976;Melrose et al., 2008;Schollmeier et al., 2000). Elastin in the AF is concentrated between adjacent lamellae and helps elastic recoil following large deformations (Melrose and Ghosh, 1988;Yu et al., 2007Yu et al., , 2005.

Cartilaginous endplate (CEP)
The CEPs are thin layers of hyaline cartilage that bind the disc inferiorly and superiorly to the adjacent bony endplates (BEPs). They are approximately 0.6 mm thick, although they are generally thinner towards the centre where they are in contact with the NP (Roberts et al., , 1989Vernon-Roberts and Pirie, 1977). Thickness decreases with age. Dimensions in the transverse plane reflect those of the adjacent vertebral bodies, and so increase from an anteriorposterior length of 16-19 mm and lateral width of 17-29 mm in the cervical spine (Francis, 1955;Panjabi et al., 1991), rising to an anterior-posterior length of 30-36 mm and lateral width of 43-54 mm in the lumbar spine (Berry et al., 1987;Panjabi et al., 1992;Scoles et al., 1988). The CEP comprises approximately 60% water, and major dry weight constituents are collagen type II and proteoglycans. Collagen content is higher, and proteoglycan content lower, near the disc periphery (Roberts et al., 1989). The three-dimensional type II collagen network of the CEP prohibits swelling, and the relatively stable tissue is able to reduce the rate of water expulsion from the pressurised NP, while allowing nutrients to diffuse into the disc from the vertebral body (Roberts et al., 1989). The CEP is bonded weakly to a thin underlying layer of perforated cortical bone, the BEP. Under compressive loading, the pressure of the NP pressing against the CEP and BEP can cause them to bulge into the VB by up to 1 mm (Brinckmann et al., 1983;Holmes et al., 1993). This bulging increases the volume available to the NP, thereby reducing pressure in the nucleus and shifting some compressive load-bearing from the NP to the AF (Adams, 2015).

Integration between components of the IVD
Fibres of the outer AF are deeply anchored into the BEP, while fibres of the inner AF merge gradually into the CEP (Hashizume, 1980;Inoue, 1981;Rodrigues et al., 2012). The outer margin of the AF also merges into the anterior and posterior longitudinal ligaments, which are considered by some to be peripheral parts of the AF (Bogduk, 2005;Coventry et al., 1945). The NP and inner AF are connected by branches of fibres that blend the boundary between the NP and AF providing mechanical integration (Wade et al., 2012a). Fine collagen type II fibrils of the NP also insert deep into the CEP (Wade et al., 2012b).

Specimen harvesting and storage
Ideally, specimens will be harvested fresh immediately prior to testing. It is, however, common practise for specimens to be deep frozen for various amounts of time before testing. Generally, a single freeze-thaw cycle has little effect on the pressure in the NP, the stiffness or creep behaviour of the IVD (Dhillon et al., 2001;Nachemson, 1960;Panjabi et al., 1985;Smeathers and Joanes, 1988), and on the tensile behaviour of small samples of AF (Galante, 1967). After several freezethaw cycles, however, significant differences in joint flexibility can appear (Tan and Uppuganti, 2012). A number of animal studies have found significant mechanical differences between fresh and freshfrozen IVDs (Bass et al., 1997;Callaghan and McGill, 1995;Sunni et al., 2014), but this may be attributable to the very high water content of young animal IVDs (Bass et al., 1997). To the authors' knowledge, the effects of precise freezing temperature and duration on the behaviour of the IVD have not been investigated.

Testing environment
Temperature and hydration should be controlled to replicate the physiological environment. At 37°C, compressive creep is approximately 10% more than at room temperature (Koeller et al., 1986), and immersion in fluid reduces an IVD's stiffness in torsion, axial compression, and lateral bending by~20-30% (Costi et al., 2002), possibly because of changes in inter-lamellar friction and/or hydraulic permeability. Specimen hydration during testing can be maintained by: 1) spraying water at regular intervals, 2) wrapping in saline soaked gauze, 3) testing in a humidity chamber, or 4) immersing in a saline bath. The last technique can cause problems, because cadaveric IVDs will swell by 20% if stored unloaded in a wet environment (McMillan et al., 1996), despite tension in the ligamentum flavum, which generates a pressure in the NP of approximately 70 kPa (Heuer et al., 2007;Nachemson and Evans, 1968). In vivo, there is a corresponding diurnal variation in hydration (and height) of the IVD of approximately 20% (Botsford et al., 1994). To allow for this, cadaveric specimens are often creeploaded prior to testing to return the hydration of the IVD to within the physiological range (Adams, 1995;Pflaster et al., 1997;Race et al., 2000). Phosphate-buffered saline (PBS) with an osmolarity of 0.15 M achieves a stable tissue hydration ex vivo (Ebara et al., 1996), with IVD compressive stiffness increasing at higher osmolarity (Bezci et al., 2015).

Strain rate, preload, and preconditioning
Strain rate affects the mechanical behaviour of the viscoelastic IVD (Holzapfel et al., 2005). Generally, the quicker an IVD is compressed or rotated, the stiffer it becomes Kemper et al., 2007;Smeathers and Joanes, 1988;Yingling et al., 1997). The non-linear properties of a disc mean that stiffness also increases if a compressive preload is applied prior to another mode of loading (Janevic et al., 1991), so it is usual to apply a compressive preload to simulate superincumbent body weight during testing. In multisegmental specimenswhich are prone to bucklinga preload of up to 1.2 kN (depending on the segments that are being tested) is often applied as a Fig. 3. Typical non-linear behaviour of a VB-disc-VB specimen subjected to uniaxial compression. The difference between the loading and unloading paths indicates viscoelastic properties (Asano et al., 1992;Koeller et al., 1986). Table 1 Overview of human axial IVD compression tests. Where numerical values were lacking, estimates have been taken from figures. Where healthy and degenerate discs have been tested, data from the healthy discs are recorded. Where there was more than one level of preload, data refer to the largest. Incremental loading refers to forces applied in discrete steps. *Only specimens tested in axial compression are included. **Maximum loads and displacements are taken from the static compression tests, not the fatigue tests. ***A compressive preload was only applied for the dynamic tests in this study. Maximum load and displacement are taken from static tests. Minimum age was between 0 and 10 years but 10 years was reported here +Preconditioned to a displacement of 0.5 mm rather than a load, therefore data not presented. ++Compressive preload of 0.4 MPa.    'follower load' along a path tangent to the curve of the spine (Patwardhan et al., 2003(Patwardhan et al., , 2000(Patwardhan et al., , 1999Stanley et al., 2004). Preconditioning, meaning cyclic loading before the intended loading protocol is applied, alleviates the effects of freezing and prolonged immobility and so ensures reproducibility. Generally, three precycles are sufficient for the response of the IVD to be consistent (Wilke et al., 1998), although some studies have incorporated thousands of preloading cycles (Costi et al., 2014;Wilke et al., 2013).

Disc age and degeneration
Age-related biochemical changes stiffen cartilage at a materials level (Bank et al., 1998), and under shear and confined compression a degenerating AF becomes stiffer (Iatridis et al., 1999(Iatridis et al., , 1998O'Connell et al., 2009). In tension, however, the accumulation of small structural defects with degeneration probably explains why AF becomes softer and weaker (Shan et al., 2015). The proteoglycan and water concentration in the NP reduces in degenerate IVDs (Buckwalter, 1995), causing the nucleus pressure to fall Sato et al., 1999). The sagittal diameter of the NP reduces by approximately 50%, and the NP begins to act more like a solid than a fluid Iatridis et al., 1997a;Johannessen and Elliott, 2005) so that more compressive load-bearing is transferred to the AF Nachemson, 1965). With increasing age, calcification of the CEP occurs, which may contribute to disc degeneration by reducing permeability and metabolite transport (Bernick and Cailliet, 1982;Nachemson et al., 1970;Roberts et al., 1996). A degenerated disc has lost height, and intervertebral ligaments have become slack (Adams et al., 1987), so the bending stiffness of motion segments decreases with level of degeneration, and therefore age (Moroney et al., 1988;Nachemson et al., 1979;Yoganandan et al., 1989;Zhao et al., 2005). In addition, their range of motion decreases, possibly because of osteophyte growth (Al-Rawahi et al., 2011). Such profound changes in the biomechanics of the IVD render assessing the level of disc degeneration before mechanical testing an important element of the experimental protocol. Published criteria allow IVD degeneration to be graded on the basis of MRI (Johannessen et al., 2006;Pfirrmann et al., 2001;Schneiderman et al., 1987), radiographic appearance (Gordon et al., 1991), or simply from visual appearance of IVDs sectioned in the sagittal or transverse plane (Galante, 1967;Thompson et al., 1990).

Spinal level
The size and shape of IVDs vary with spinal level (Pooni et al., 1986), giving rise to systematic changes in intradiscal stresses (Dolan et al., 2013). After normalisation for size, cervical motion segments are stronger in compression but weaker in bending compared to lumbar (Przybyla et al., 2007). Lumbar IVDs have been the focus of most investigations concerned with IVD degeneration or low back pain. The mechanical responses of the five lumbar IVDs are similar in extension, lateral bending, compression, and shear , although the thicker lower lumbar discs lose more height under creep loading (Hirsch and Nachemson, 1954), and show greater anterior bulging during axial dynamic compression (Koeller et al., 1984a). Torsional stiffness of whole motion segments increases at lumbar and lower thoracic levels because of increased vertebral size, and altered orientation of the zygapophysial joints in comparison to upper thoracic and cervical levels (Markolf, 1972;Panjabi et al., 1993).

Biomechanical testing of functional spinal units
Tests usually involve a functional spinal unit (FSU)or motion segmentwhich comprises an intact IVD still connected to its two adjacent vertebrae, with all ligaments intact. The posterior 'neural arch' region of each vertebra articulates with adjacent neural arches by means of two sliding synovial joints, the zygapophysial joints, and these  also are preserved in a motion segment (Fig. 1). If it is required to distinguish the properties of the IVD itself from those of adjacent ligaments and zygapophysial joints, then the neural arches can be removed at the pedicles, leaving a VB-disc-VB specimen. Large specimens comprising three or more vertebrae, with all posterior structures intact, are sometimes tested because they allow adjacent-level effects to be studied, or large prostheses or fixations to be evaluated. As the focus of this review is the IVD, these experiments will not be included here. Some studies have loaded the FSU to failure in order to determine its strength for a particular mode of loading, whereas others have applied sub-failure loads in order to simulate physiological conditions. Although the two approaches have methodological differences, both are important to characterise different aspects of the biomechanical behaviour of the IVD. Although the IVDs normally carry most of the compressive load, more than 50% can be transmitted through the neural arch of the joint complex (Pollintine et al., 2004). Load bearing by the neural arch is greater when the spine is extended than when it is flexed Asano et al., 1992;Hirsch and Nachemson, 1954;Lorenz et al., 1983;Shirazi-Adl and Drouin, 1987;Yang and King, 1984) and is increased when the IVDs are narrowed, either by sustained loading or by degenerative changes (Pollintine et al., 2004). To ensure that mechanical loading is applied appropriately to both anterior (vertebral bodies and IVDs) and posterior (neural arch) columns, it is important for motion segments to be secured in some stiff fixative such as dental cement before testing.
Motion segments have been tested under uniaxial compression, axial rotation, lateral bending, and flexion/extension. The following sections are categorised according to these loading modes.

Bending and axial rotation
Because spinal joints are subjected to complex three-dimensional loading in vivo, a number of mechanical spine testers have been devised which are capable of applying combinations of compression, bending, and torsion to spinal segments ex vivo. Typical momentrotation graphs (Fig. 4) show marked nonlinearity and hysteresis, and do not necessarily start at the origin because of pre-conditioning. Generally, resistance is greatest in axial rotation, followed by extension (backwards bending), lateral bending, and finally flexion (forwards bending) (Beaubien et al., 2005;Moroney et al., 1988;Schultz et al., 1979).
One key decision when applying moments to spinal joints is whether to rotate about a fixed axis, or allow the rotation axis to move and thus minimise shear. These two strategies are often described as being equivalent to 'displacement control' or 'moment control', respectively. Moments (or displacements) can be applied in a quasi-static manner by means of suspended dead-weights, or dynamically either using servo-hydraulic or electric actuators that apply moments via cables or by offsetting a compressive load. Table 2 summarises previous spinal flexion-extension, lateral bending, and axial rotation tests performed on single motion segments. Note that a number of the studies included in Table 2 have combined different modes of loading, for example Adams and Hutton (1981) combined compression and torsion, Gordon et al. (1991) combined flexion, rotation and compression, and  combined compression and bending. Although not reported in Table 2, in order to reduce viscoelastic effects, specimens are often preconditioned for a number of cyclesusually 3until a reproducible result is achieved.

IVD injuries
Compressive overload always damages the BEP first, and vertical herniation of the NP can create Schmorl's nodes (Adams and Dolan, 2012;Brinckmann et al., 1988;Hamanishi et al., 1994). The resulting decompression of the NP may lead to internal collapse of the AF (Adams and Dolan, 2012;Gunzburg et al., 1992). For herniation to occur through the AF the IVD needs to be flexed or overloaded in  Table 2 Summary of human flexion-extension, lateral bending, and axial rotation tests on single motion segments. If a test involved more than one level of compressive preload, data refer to the largest. *Average maximum moments and rotations are not presented as the loading protocol included a combination of compression, flexion, bending or rotation. +In this study the lamina were removed but the zygapophysial joints were left intact. multiple modes simultaneously; the likelihood of herniation to occur under load is higher at high loading rates Wade et al., 2014Wade et al., , 2015. Excessive torsion damages the neural arch (Adams and Hutton, 1981), and may cause delamination of the outer anulus (Farfan et al., 1970). Complex loading in bending and compression can lead to radial fissures (Adams and Hutton, 1983), and disc prolapse ('slipped disc'), either in a single loading cycle , or by fatigue failure (Adams and Hutton, 1985). Axial rotation probably enhances the vulnerability of the posterior portions of the IVD to damage, and reduces the load required to cause disc failure (Callaghan and McGill, 2001;Drake et al., 2005;Gordon et al., 1991;Pearcy and Hindle, 1991;Shirazi-Adl et al., 1986;Veres et al., 2010).

Material properties of IVD constituent tissue
To gain a full understanding of the mechanical behaviour of the IVD as a whole, it is important to understand the behaviour of its component tissues.

Nucleus pulposus (NP)
Material properties of the NP are poorly characterised, reflecting the fact that it is extremely difficult to test in isolation. There is even debate as to whether it behaves predominantly as a solid or a liquid, with solid behaviour being more apparent under dynamic conditions . In tension, the tissue is so non-linear that it has been likened to a 'tethered fluid' (Skrzypiec et al., 2007a;Wade et al., 2012a). NP properties have been measured in confined compression (Johannessen and Elliott, 2005;Yang and Kish, 1988), unconfined compression (Cloyd et al., 2007), and shear (Bodine et al., 1982;Iatridis et al., 1997b). In confined compression, reported effective modulus of human, non-degenerate NP is 1.0 MPa (Johannessen and Elliott, 2005), and bulk modulus is 1720 MPa (Yang and Kish, 1988). Although the behaviour of the NP is not linearly elastic, Cloyd et al. (2007) reported a modulus from the 'linear' region of the stress-strain response in unconfined compression of 5.4 kPa. Shear moduli have been reported between 7 and 50 kPa (Bodine et al., 1982;Iatridis et al., 1997b). Table 3 summarises mechanical characterisation studies of the human AF according to test protocol, type of specimen, loading rate, direction of loading, and method of gripping. Galante (1967) also tested multiple lamina specimens at various angles, including across fibre, however these results have not been included in Table 3 in order to avoid confusion with results obtained from single layer cross-fibre and in-line-with-fibre specimens.

Anulus fibrosus (AF)
A number of studies in Table 3 found that the outer and anterior regions of the AF were considerably stiffer than the inner and posterior regions of the AF (Ebara et al., 1996;Elliott and Setton, 2001;Holzapfel et al., 2005;Shan et al., 2015;Skaggs et al., 1994). In tension, the AF is stiffest along the axis of the fibres (average modulus=183 MPa), followed by the circumferential direction (16 MPa), and finally the radial, axial and cross fibre directions (where average moduli were 0.3, 2.6 and 0.2 MPa, respectively). In circumferential tension the average failure stress was 2.7 MPa and the average failure strain was 0.3. The maximum strain rate at which the tensile behaviour of the AF has been investigated is 4%/s which may be considered to be relatively low compared to the strain rates expected in many physiological activities and indeed in injurious scenarios. The tensile properties of small anulus specimens also depend on specimen size, because small excised specimens sustain proportionally greater disruption to the collagen network .
Gripping small excised specimens of soft tissue is a challenge,   Table 3 Overview of tensile tests on the human AF. Where specimens from different regions of the AF were included, an average is recorded. Where numerical data were not presented, values have been estimated from figures. *Circumferential failure stress/strain is reported here, the along fibre (multiple lamella) failure stress was higher at 8.8 MPa and the failure strain was lower at 0.25. +Along fibre, rather than radial failure stress and failure strain is reported here, radial failure stress was lower and failure strain was higher (0.187 MPa and 1.61).

Results
Strain rate (%/s)   Shan et al. (2015) particularly in uniaxial tension tests where slippage is common. Loaded clamps are often used, particularly in combination with sandpaper or polishing paper, and cyanoacrylate glue (Fujita et al., 2000(Fujita et al., , 1997Shan et al., 2015;Skaggs et al., 1994;Skrzypiec et al., 2007b;Smith et al., 2008). Tungsten rakes, wire rods pierced through the specimen, and aluminium rings clamped around the specimen have also been used (Bass et al., 2004;Wagner and Lotz, 2004). Slippage can give rise to large errors if specimen strain is calculated from displacement of the clamps. Measuring strain optically eliminates this problem, and enables regional variations in strain to be studied (Masouros et al., 2009). Diagrammatic representations of common gripping methods are presented in Fig. 5.

Cartilaginous (CEP) and bony (BEP) endplates
The CEP has not been studied extensively either, possibly because the tissue layer is so thin rendering it difficult to harvest. CEP fragments often appear in IVD herniations, which may be because the CEP-vertebra junction is very weak or because the CEP-NP integration is relatively strong (Lama et al., 2014;Wade et al., 2014). CEP strength has been characterised by a number of studies in order to understand failure mechanisms and to help design interbody implants or grafts (Grant et al., 2001;Perey, 1957). Although the properties of the CEP are not linearly elastic, a modulus value has been estimated as 23.8 MPa (Yamada, 1970).
Indentation tests show that the underlying BEP is stiffest and strongest around the periphery of the VB, with stiffness of healthy BEPs in the region of 75-175 N/mm (Grant et al., 2001;Liu et al., 2016). Stiffness and strength both decrease in the presence of IVD degeneration (Liu et al., 2016). Deflection of whole lumbar endplates BEPs has also been studied during compression of VB-disc-VB specimens where, at the centre of the caudal endplate; the compliance is approximately 0.1 mm/kN (Holmes et al., 1993). Failure is more common in the weaker cranial BEP, because it is 14% thinner than the caudal BEP and supported by less dense trabecular bone (Zhao et al., 2009).

Discussion
This paper has provided a detailed review of the techniques used, and results obtained by, studies that have attempted to characterise mechanically the IVD. The IVD is a complex structure which changes quite markedly with age and degeneration, and IVD dysfunction can adversely affect quality of life. Disc mechanical responses are sensitive to the testing environment and boundary conditions applied, so care must be taken to ensure that applied loading is appropriate for each specific research question. A wide range of testing techniques have been used, but a consensus is emerging regarding such details as the need to precondition specimens to obtain a more repeatable response during testing, and the need to take account of posture, loading rate, and disc age and degeneration when comparing results between experiments. Future experimental studies will benefit from finding a similar consensus regarding testing environment (humidity, temperature) and methods of pre-test storage. We suggest that future work is required in the following specific research areas: 1) the influence of spinal level on the mechanical properties of constitutive components of the motion segment; 2) the responses of the IVD to complex (combined) loading modes, particularly in relation to injury mechanisms; 3) mechanical properties of the nucleus pulposus at varying strain rates and levels of degeneration; and 4) high strain-rate material properties of the anulus fibrosus.
It was apparent when performing this review that breakthroughs in our understanding of the biomechanics of the IVD often coincided with the introduction of a new measuring capability or technology. For example, new techniques for imaging the disc, for applying complex loading to cadaveric specimens, and for quantifying stress and strain distributions within discs have contributed greatly to our understanding of how discs function and fail. This effect is likely to continue into the future, and it emphasises the importance of understanding variations in past techniques in order to develop new ones.