|Sciatica: Studies of symptoms, genetic factors, and treatment with periradicular infiltration|
|Prev||Chapter 2. Review of the literature||Next|
Traditionally, compression of nerve roots or DRGs by the HNP has been regarded as the cause of sciatica, although HNP can be found in 20% to 36%, depending on the age, of asymptomatic subjects (Wiesel et al. 1984, Boden et al. 1990, Jensen et al. 1994). Similarly, internal disc ruptures (without HNP) may also induce disabling radicular pain, indicating the existence of an alternative mechanism to neural compression (Ohnmeiss et al. 1997, Ohnmeiss et al. 1999).
The effect of compression on the nerve roots has been studied extensively in animal models. Rapid onset (0.05–0.1 s) compression of porcine nerve roots induced intraneural oedema after 2 minutes compression at 50 mmHg, whereas following slow onset (15–20 s) compression, oedema occurred after 2 hours (Olmarker 1991). Significant reduction (20–30 %) of nutrition was already observed at 10 mmHg compression, probably due to impairment of cerebrospinal fluid flow (Olmarker 1991). Intraneural oedema may increase the endoneural fluid pressure and lead to impairment of intraneural blood flow as in the closed compartment syndrome (Rydevik et al. 1989). In dogs, compression induced a marked extravasation of protein tracers; electron microscopy showed the tight junctions of the intraneural capillaries opened and vesicular transport increased, indicating disruption of the blood-nerve barrier (Kobayashi et al. 1993). In MRI, this was reflected as enhancement of the compressed nerve roots.
When normal feline dorsal roots were compressed with chromic gut ligatures, only a brief discharge of maximally 20 impulses was observed (Howe et al. 1977). Chronic injury, however, greatly increased their mechanical sensitivity. When chronic compression of canine nerve roots was studied with a silastic tube model, impairment of the blood-nerve barrier (thickening of the dura and arachnoid membrane around the affected nerve root) was observed after 1 month, suggesting that the most important factor in nerve root dysfunction due to chronic compression is intraradicular oedema induced by increased local vascular permeability (Yoshizawa et al. 1995). Similar nerve root enhancement and axonal degeneration, along with marked inflammatory cell infiltration, were also found in baboons (Nguyen et al. 1995). Destruction of large myelinated fibers was observed after 1 week of gradual chronic compression of porcine nerve roots (Cornefjord et al. 1997). Endoneural bleeding and signs of inflammation in the compressed nerve roots were more common after 1 than 4 weeks.
In contrast to nerve roots, even normal noninjured DRGs were very sensitive to compression, gentle pressure producing prolonged repetitive firing in single afferent fibers (Howe et al. 1977). This was confirmed by measuring the responses in dorsal horn wide dynamic range neurons to compression of dorsal root or DRG (Hanai et al. 1996). Compression of the DRG, but not the dorsal root, produced prolonged repetitive firing. Hypoxia further increased sensitivity to mechanical stimuli and even evoked spontaneous firing in an in vitro model (Sugawara et al. 1996). The relevance of DRGs in the pathophysiology of sciatica is supported by observations that the DRG is the most likely site of compression from a herniated disc (Lindblom & Rexed 1948, Rydevik et al. 1984).
The extruded nuclear material of the disc is chemically inflammatory and neurotoxic (McCarron et al. 1987). When in contact with the nerve roots, the nuclear material (without any compression) induces structural and functional changes in porcine nerve roots (Olmarker et al. 1993). The functional changes include focal degeneration of myelinated fibers and focal Schwann cell damage in nondegenerated axons. The damage to the Schwann cells resulted in a disintegration of Schmidt-Lanterman incisures (Olmarker et al. 1996). These incisures connect the cytoplasm of the Schwann cells situated outside the myelin sheath to the part on its inner side, the external parts of the Schwann cells being essential for the normal impulse conduction properties of the axons. These studies were performed by applying a large amount of NP on the nerve roots, but similar functional nerve root damage was also observed in pigs after experimental disc herniation (Kayama et al. 1996).
The nerve fiber damage induced by the NP could be due to its toxicity. In fact, many substances, including hydrogen ions and glycoproteins, has been suspected to cause chemical radiculitis (Nachemson 1969, Marshall et al. 1977). However, NP is also chemotactic, attracting leukocytes, and it may also induce macromolecular leakage and spontaneous firing of axons in vitro (Olmarker et al. 1995). In chronic compression studies, inflammatory cell infiltrates, mainly macrophages, have been observed (Yoshizawa et al. 1995, Nguyen et al. 1995). Furthermore, in a rat model of mechanical hyperalgesia induced by application of NP to nerve roots, depletion of leukocytes inhibited the generation of hyperalgesia (Kawakami et al. 2000b). This indicated that leukotactic properties of the HNP are important in the production of pain-related behaviour. The cells first appearing in and around the herniated NP on nerve root were polymorphonuclear leukocytes, whereas macrophages, originating from monocytes, did not predominate until after a few days and then remained in the affected region until the inflammation subsided (Kawakami et al. 2000b).
The observed decrease in nerve conduction velocity may be due to impaired ion exchange following changes in Schmidt-Lanterman incisures (Olmarker et al. 1996) or to ischaemia (Kayama et al. 1996). The latter mechanism is supported by the finding that autologous NP increased endoneural pressure and reduced blood flow (assessed with a laser Doppler flow probe) by 10 % to 20 % in the DRG (Yabuki et al. 1998a). Irritation of a nerve root thus caused a “compartment syndrome” in the DRG. Interestingly, blood flow in the ipsilateral hind paw was reduced, too (Yabuki et al. 2000). Also after experimental herniation, blood flow in canine nerve root was reduced, correlating with the decrease in nerve conduction velocity (Otani et al. 1999). The decrease in blood flow was maximal at 1 week, recovering within 1 month. In the DRG, however, there was a statistically significant decrease in blood flow only at 1 week, and no reduction/recovery pattern was observed (Otani et al. 1999).
The neurotoxicity of the NP seems to be associated with disc cells, as freezing prevented the neuronal damage (Olmarker et al. 1997, Kayama et al. 1998). In the following, the possible inflammatory candidates in the NP are discussed.
Phospholipase A2. Phospholipase A2 (PLA2) is the rate-limiting enzyme in the synthesis of proinflammatory lipid mediators (prostaglandins, leukotrienes, lipoxenies, and platelet-activating factor). The enzyme liberates arachidonic acid from the membrane phospholipids, and is secreted extracellularly by activated phagocytes in response to cytokines (Vadas & Pruzanski 1986). Additionally, it is released from rabbit chondrocytes by IL-1 (Chang et al. 1986). PLA2 is found in extraordinarily high concentrations in herniated and painful discs (Saal et al. 1990b), and the enzyme is also itself inflammatory (Franson et al. 1992). PLA2 is calcium-dependent, adsorbing tightly to plasma membranes and intact cells (Vadas & Pruzanski 1986).
When PLA2 was injected into the nerve receptive fields of isolated rabbit facet joints, it produced sensitization of the nerves and recruitment of ”silent neural units”. Histologically, inflammation was observed in the samples 2 hours after the injection (Özaktay et al. 1998). This resembled the observed neuroexitatory effect of NP (multiunit discharge lasting for several minutes) (Cavanaugh 1995). Furthermore, when PLA2 was injected epidurally, motor weakness, demyelinisation, and increased sensitivity of dorsal roots to mechanical stimulation were observed after 3 days, but not beyond 3 weeks (Chen et al. 1997). This could explain the finding of low phospholipase activities in herniated tissue samples (Grönblad et al. 1996), because surgery is usually undertaken several weeks after the onset of symptoms. In fact, phospholipase A2 activity was found to be maximal 1 week after chromic gut ligature, whereas thermal hyperalgesia was maximal 3 weeks after surgery (Lee et al. 1998). In a rat model, NP-induced mechanical hyperalgesia was abolished by mepacrine, a selective inhibitor of phospholipase A2 (Kawakami et al. 1998). Interestingly, anulus or a combination of anulus and NP induced mechanical hyperalgesia only after addition of a selective inhibitor of inducible nitric oxide synthase (iNOS), indicating a role for nitric oxide (NO) in reducing mechanical hyperalgesia in this model (Kawakami et al. 1998).
Tumor necrosis factor (TNF-a). TNF-α is a cytokine produced mainly by activated macrophages and T cells in response to inflammation, and by mast cells and Schwann cells in response to peripheral nerve injury (Wagner & Myers 1996b, Bemelmans et al. 1996). It activates the transcription factors NF-κ B and AP-1 by binding to its p55 TNF receptor (TNFRI), thereby inducing the production of proinflammatory and immunomodulatory genes (Darnay & Aggarwal 1997). Endoneurial TNF-α causes demyelinisation, axonal degeneration, and hyperalgesic pain states (Wagner & Myers 1996a), while anti-inflammatory IL-10 treatment in chronic constriction injury to peripheral nerves decreases thermal hyperalgesia, macrophage recruitment and endoneurial TNF-α expression (Wagner et al. 1998). However, TNF-α also has an important role in the resorption of disc herniation (Haro et al. 2000). Macrophages secrete matrilysin-enzyme (MMP-7), which liberates soluble TNF-α from macrophage cell membranes. Soluble TNF-α induced disc chondrocytes to secrete stromelysin-1 enzyme (MMP-3), which was required for the release of a macrophage chemoattractant and subsequent macrophage infiltration of the disc (Haro et al. 2000).
In thermal hyperalgesia, two peaks have been associated with Wallerian degeneration, and can be reproduced in chronic injury to peripheral nerves (Shubayev & Myers 2000). These peaks are also related to changes in TNF-α expression. It seems that the first TNF-α peak, 6 hours after the peripheral nerve injury, is due to the local expression of the cytotoxic transmembrane 26 kDa TNF-α protein released by the resident Schwann cell, mast cells and macrophages. This peak in TNF-α expression corresponds to an increase in activity of gelatinase B (MMP-9), which is already greatly upregulated 3 hours after the injury. The second peak occurs 5 days after the injury, and may represent TNF-α protein released by haematogenously recruited macrophages. The second peak corresponds to an increase in soluble 17 kDa TNF-α and gelatinase A (MMP-2) upregulation (Shubayev & Myers 2000). When degenerated and normal human articular cartilage were compared, TNF-α (and IL-1β ) was expressed only in the superficial zone of degenerated cartilage. However, the phenotype of chondrocytes varied widely even in degenerated cartilage, 5 to 40 % of cells expressing these cytokines (Tetlow et al. 2001).
Recently it was found immunohistochemically that TNF-α was expressed in the NP (Olmarker & Larsson 1998); in fact, 17 kDa cytokine was expressed at a concentration of approximately 0.5 ng per disc (Igarashi et al. 2000). Consistent with the results of Haro and co-workers (Haro et al. 2000), TNF-α (and other cytokines) was produced in protrusion type herniations by chondrocytes, but in extrusions by histiocytes, fibroblasts, and endothelial cells constituting granulation tissue (Takahashi et al. 1996).
Exogenous TNF-α produced neuropathological and behavioural changes (Wallerian degeneration of nerve fibers, macrophage recruitment to phagocytoze the debris, splitting of the myelin sheath) that mimicked those of the NP (Igarashi et al. 2000). Later, in a chronic peripheral nerve injury model, remyelinisation and reactive changes in endothelial cells (collagen deposition in response to fibroblast activation) have been observed after intraneural injection of TNF-α (Redford et al. 1995). In herniated disc tissues similar changes, such as endothelial proliferation, vascular activation and collagen proliferation have been observed (Cooper et al. 1995). Treatment with doxycycline, a nonspecific TNF-α antagonist, blocked the NP-induced reduction of nerve conduction velocity (Olmarker & Larsson 1998). More recently, both soluble TNF-α receptor (Embrel™) and TNF-α antibodies (Remicade™) reversed NP-induced nerve conduction block (Olmarker & Rydevik 2001). Furthermore, these antagonists also specifically blocked the NP-induced oedema and thrombus formation, indicating that these vascular changes were TNF-α mediated. In another study, topical pentoxifylline, an inhibitor of TNF-α synthesis, prevented the NP-induced compartment syndrome in the DRG (Yabuki et al. 2001). In conclusion, it seems that TNF-α exerts a crucial role in NP-induced nerve root damage.
Other inflammatory mediators. IL-1 plays an important role in experimental allergic radiculitis induced in rats, since IL-1 receptor antagonist ameliorated the symptoms (Wehling et al. 1996). In fact, IL-1 and IFNγ act synergistically with TNF-α and are more or less neurotoxic (Chao et al. 1995). In vitro, herniated discs spontaneously produce nitric oxide, matrix metalloproteinases, IL-6 and PgE2 (Kang et al. 1997). Even control discs synthetized these substances when the tissue samples were exposed to IL-1β (but not without). In herniated discs, IL-1β further increased the production of NO, IL-6 and PgE2 (Kang et al. 1997).
In a canine model, PgE2 produced ectopic firing of nerve roots, which was suppressed with steroid (Muramoto et al. 1997). PgE2 production in vitro could also be blocked with inducible cyclo-oxygenase enzyme (COX-2) inhibitor (Miyamoto et al. 2000).
NO seems to play an important role in radicular pain. In a disc herniation sample, iNOS was produced by cells in the granulation tissue around the extruded anulus fibrosus (Hashizume et al. 1997). Autologous epidurally applied AF (but not NP) produced thermal hyperalgesia in a rat model (Kawakami et al. 1998). Thermal hyperalgesia was abated with epidural saline and abolished with a specific inhibitor of iNOS. In pigs, a specific inhibitor of iNOS, aminoguanidine, prevented the formation of NP-induced oedema and a negative effect on the nerve conduction velocity (Brisby et al. 2000), but in the rat model NO reduced mechanical hyperalgesia (Kawakami et al. 1998). NO seems thus to have a dual action on the nerve roots, both negative and positive, similar to articular cartilage where it is involved in both the catabolism and synthesis of PGs (Stefanovic-Racic et al. 1996).
When the effects of the NP, compression by silk ligature, and the combination of these two were compared in rats, thermal hyperalgesia was induced only by the combination of compression and NP (Kawakami et al. 2000a). Even though silk ligature did not effect thermal withdrawal latency, histologically there were fewer large and more small-diameter fibers than with the combination of silk and NP (Kawakami et al. 2000a). Similar results were observed in a study where three experimental procedures (either NP applied on the rat L4 nerve root, nerve root displacement with a needle, or a combination) were compared in a rat model (Olmarker & Myers 1998). There was a significant reduction in mechanical threshold at days 2, 4, 16 and 18, and significant thermal hyperalgesia from day 2 until day 14 post-operatively in the combination protocol animals. Histologically, a significant cellular injury by 21 days post-operatively was noticeable. Oedema and fibrotic reactions were observed in the subperitoneal/root sheath area and in the endoneural space. There was also perivascular oedema and indications of reactive endothelial cells, axonal demyelination, myelin splitting, and Schwann cell hypertrophy (Olmarker & Myers 1998). On the basis of these studies, the probable scenario is that the ruptured intervertebral disc with leaking nucleus sensitizes the nerve root(s), and in the presence of mechanical deformation, pain-related behaviour is induced. This accords with observations that stimulation of nerve root in contact with herniated disc tissue reproduces sciatic pain (Smyth & Wright 1958, Kuslich et al. 1991). In these animal models, autologous NP has mainly been used. However, in articular cartilage, TNF-α expression was confined only to the degenerated samples, and even in degenerated cartilage 5–40 % of cells expressed the cytokine (Tetlow et al. 2001). Nondegenerated animal disc tissue may, thus, not be fully comparable with degenerated tissue. This explains the discrepancy between these animal results and the clinical situation where prolonged sciatica is encountered without disc herniation.
Peripheral nerve endings become sensitized by chemical mediators that are released during tissue damage and inflammation. These include neurogenic mediators, such as substance P, and non-neurogenic mediators, such as bradykinin, histamine and prostaglandins (Cavanaugh 1995). In addition to local increase in the excitability of C fibers, spinal mechanisms are important in the production and maintenance of hyperalgesia. Thermal hyperalgesia requires the activation of N-methyl-D-aspartate (NMDA) receptors and is primarily mediated by production of nitric oxide, whereas mechanical hyperalgesia requires α-amino-3-hydroxy-5-methylisoxazole-5-propionate (AMPA) and metabotropic glutamate receptor coactivation, and is primarily mediated by cyclooxygenase products and PLA2 activation (Meller & Gebhart 1994).
In rats, compression of nerve roots by loose chromic gut ligatures induced prolonged thermal hyperalgesia (related to neuropathic pain), and initial transient motor dysfunction and mechanical hypoalgesia (Kawakami et al. 1994a). Pain-related behaviour correlated with increased dorsal horn c-fos and dorsal ganglion VIP amounts. Similarly, DRG irritation generated thermal hyperalgesia, which was accompanied by increased c-fos expression and spontaneous pain-related behaviour (Chatani et al. 1995).
Epidural betamethasone, but not bupivacaine, inhibited thermal hyperalgesia generated by ligating rat nerve roots with chromic gut ligatures (Hayashi et al. 1998). However, in this model the positive effect of steroid did not correlate with the changes in SP, CGRP, and c-fos expression. Histological nerve fiber damage did not correlate with pain-related behaviour, which indicates that rather than mechanical compression, a chemical irritant released from chromic gut is responsible for thermal hyperalgesia (Kawakami et al. 1994b). Interestingly, in the rat model PLA2 was involved in mechanical hyperalgesia induced by the NP, and NO in thermal hyperalgesia induced by the AF (Kawakami et al. 1998).
In a rat model, astrocytic activation demonstrated a direct relationship with the mechanical allodynia for the first 7 days. Simultaneously, spinal cord IL-1β secretion was increased, indicating a neuroimmune component in lumbar radiculopathy (Hashizume et al. 2000). Similarly, one week after relocation of autologous NP on rat L4 and L5 nerve roots, nerve root and DRG IL-1β expression were increased. In the same study, IL-6 expression was observed in these tissues over the whole 4-week period (Kawakami et al. 1999).
Methylprednisolone injected within 48 hours after the application of the NP inhibited the NP-induced vascular permeability and functional impairment (decrease of nerve conduction velocity) (Olmarker et al. 1994). Histologically, no differences between the NP and NP + methylprednisolone groups were observed. The nerve loss was, however, focal (Olmarker et al. 1994). Methylprednisolone also inhibited the NP-induced increase of vascular permeability (Byröd et al. 2000).
In the rat model, PLA2 activity was found to be maximal 1 week after chromic gut ligature, whereas thermal hyperalgesia was maximal 3 weeks after surgery (Lee et al. 1998). Epidural steroid decreased PLA2 activity and reduced thermal hyperalgesia. Interestingly, steroid at different time points (1 day before, 1 day after, or 3 days after surgery) had a similar effect on thermal withdrawal latency (Lee et al. 1998).