Use of tumor necrosis factor α (TNFα)–blocking agents to treat radiculopathy associated with disc herniation is a current topic of interest (1, 2). Although the rationale for use of such treatment appears to be sound, there is no proof of its effectiveness, and its use has not been validated. In the present report, we review the evidence favoring and the evidence not favoring the involvement of TNFα in sciatica and offer insights into the use of TNF blockers as treatment.
Pathophysiology of radiculopathy with disc herniation
Since 1934, compression of the nerve root by disc herniation has been the explanation for the development of sciatica (3), with surgery to relieve the nerve compression being the reference treatment for patients in whom medical treatment has failed. However, recent evidence suggests that chemical factors could have a central role in the genesis of sciatica.
Clinical support for the mechanical theory of sciatica includes perioperative evidence (trapped nerve roots, lamination by disc herniation), a favorable outcome with surgery, and rapid postoperative improvement. However, despite the 80–90% short-term success rate of surgical treatment, the long-term success rate is 40–80% (4), with reintervention rates of 5–25% (5, 6). In addition, asymptomatic disc herniation has been reported (7), as has severe sciatica without visible disc compression (8). The severity of symptoms and the extent of the disc herniation are poorly correlated (9), and outcomes are favorable after conservative treatment and are similar with surgical or conservative treatment (6, 10).
The existence of a chemical component in the development and maintenance of sciatica has some support. Indeed, the nucleus pulposus (NP), which is isolated from the immune system after its embryologic formation, might secrete substances able to induce an autoimmune reaction in disc herniation, particularly extruded herniation (11), and immunogenicity of intervertebral discs has been proposed (11–13). The spontaneous resorption of disc herniation, as demonstrated by longitudinal computed tomography and magnetic resonance imaging (MRI) studies, appears to be more marked with voluminous or extruded herniation (14). The presence of mediators of inflammation (phospholipase A2, prostaglandin E2 [PGE2], interleukin-1α [IL-1α], IL-1β, IL-6, TNFα, and nitric oxide [NO]) in discal and peridiscal tissue was confirmed in vitro and in vivo (15–17). The mechanical compression of a healthy nerve caused dysesthesia or motor deficit but not radicular pain. In humans, mechanical stimulation of a nerve root not exposed to disc herniation produced simple discomfort, whereas stimulation of a nerve root in contact with disc herniation reproduced sciatic pain (18). In pigs, the epidural application of autologous NP, without mechanical nerve root compression, resulted in changes in nerve conduction velocity (NCV) (12), thus inducing radicular dysfunction without compression; a role of chemical substances was proposed.
However, mechanical compression plays an important role in the development of sciatica (19). The role of mechanical compression of a nerve root, that of chemical treatment, and a combination of the 2 processes have been studied (20–24). In a rat model evaluating behavioral changes related to pain, only the combination of the 2 processes produced abnormalities (20). The combination of mechanical compression and chemical irritation in dogs produced more marked electrophysiologic and histologic abnormalities (21). Application of mechanically compressed NP in rat lumbar nerve roots provoked more marked and longer-lasting hyperalgesia compared with noncompressed NP (22). Prolonged hyperalgesia was greatest, with greater intensity and earlier onset, in rats under disc compression than in those without disc compression (23). In rats, only the combination of NP treatment and displacement of the nerve root reduced the threshold of thermal stimulation and caused hyperalgesia (24). Thus, both mechanical and chemical components play a part in sciatica induced by disc herniation, with the effect being not simply additive but, rather, synergic.
Thus, even in the absence of mechanical compression, substances secreted by the NP may provoke functional and structural abnormalities of the nerve root; the pain is probably observed only when the nerve root has previously been or is simultaneously affected by a mechanical factor.
Chemical substances secreted by the NP
For more than 10 years, Olmarker and colleagues (12) investigated the chemical component in sciatica (Figure 1). Those investigators harvested NP from the L3–L4 intervertebral disc of pigs, then performed a bilateral laminectomy of the second coccygeal vertebra, exposed the cauda equina, and epidurally placed the NP gently in close contact with the spinal nerve roots; the lamina were removed, so enough room was provided for the NP to avoid mechanical compression of the cauda equine. Such treatment induced a pronounced reduction in NCV after 1–7 days (Table 1); histologic analysis showed more marked degeneration of nerve fibers with NP application (12).
Table 1. Evolution of nerve conduction velocity in 30 pigs, after application of retroperitoneal fat or nucleus pulposus*
Olmarker et al then investigated whether abnormalities observed after NP application could be reversed by treatment with antiinflammatory agents, which would provide evidence for the inflammatory nature of secreted substances. Administration of a high dose of a corticosteroid (30 mg/kg methylprednisolone) 5 minutes and 24 and 48 hours after NP application restored NCV on day 7 (25). In this model, NCV was significantly higher in animals treated with diclofenac compared with controls that received an injection with saline solution (26). In another model, incision of the L6–L7 disc reduced NCV and intraneural blood flow, and both abnormalities were reversed following treatment with indomethacin (27).
These experiments thus indicate the proinflammatory nature of the substances secreted by the NP and their ability to induce electrophysiologic changes. Olmarker and associates also investigated the effect of autologous NP that had been subjected to 1 of 3 treatments: 37°C for 24 hours, −20°C for 24 hours, or digestion by hyaluronidase for 24 hours. NCV was decreased when NP was kept at 37°C or when it was digested with hyaluronidase, but NCV was not decreased when NP was kept at −20°C (28). Because the effects of freezing were related to lysis of NP cells, these cells appeared to be at the origin of the biologic effects.
NP application can damage axons and the myelin sheath (12, 24, 25, 28, 29), increases vascular permeability (11, 30) and intravascular coagulation (11), and reduces intraneural blood flow (31), which can be inhibited by treatment with methylprednisolone (25), diclofenac (26), or indomethacin (27). Such proinflammatory properties are relatively similar to those of TNFα. Indeed, TNFα can cause nerve damage (32), particularly to myelin, similar to that observed after NP application (33), and can cause increased vascular permeability (32, 33). Such nerve damage can be inhibited by corticosteroids (34), cyclosporine (35), or nonsteroidal antiinflammatory drugs (36). Although NP application can cause behavioral modifications related to pain, particularly thermal hyperalgesia (24), TNFα also causes similar conditions and true neuropathy (32, 37). To confirm the key role of TNFα, Olmarker's model was used to investigate TNFα in NP cells and the impact of doxycycline, a powerful inhibitor of TNFα (38), and anti-TNFα monoclonal antibodies (mAb) on treating discs with NP cells (39). TNFα was identified in NP cells, and the reduced NCV after NP application was completely blocked by doxycycline and, to some extent, anti-TNFα mAb (Table 2) (39).
Table 2. Evolution of nerve conduction velocity in pigs, 3 days after application of nucleus pulposus, with or without anti-TNFα antibodies or doxycycline*
Nerve conduction velocity, msec
Values are the mean ± SD. Values from the control group in earlier experiments by Olmarker et al were 76 ± 11 msec (see refs.12 and39). TNFα = tumor necrosis factor α.
Experimental findings supporting the involvement of TNFα and the use of TNFα blockers in humans
TNFα and nerve damage.
A rat model of sciatica (chronic constriction injury [CCI]) demonstrated hyperalgesia resulting from perineural inflammation, which is similar to human peripheral neuropathies (40). Proinflammatory cytokines such as IL-1β, IL-6, and, particularly TNFα, are secreted by patients with peripheral neuropathies (41). TNFα plasma levels are increased after nerve compression (42), and endoneural injections of TNFα cause thermal hyperalgesia and mechanical allodynia, edema of the nerve root, Schwann cell damage, and activation of macrophages (32). Endogenous TNFα mediates pain behavior in models of nerve dysfunction (43), whereas exogenous TNFα causes neuronal excitation in vitro and pain in vivo (44, 45). Thalidomide, a selective and powerful inhibitor of TNFα, reduces hyperalgesia in the CCI model (46). TNFα contributes to induced neuropathic pain (46, 47); however, inhibition of TNFα and antagonism of the TNFα receptor reduce the hyperalgesia associated with sciatica (43, 48). In summary, TNFα appears to be able to sensitize the nerve root to pain induced by mechanical deformation.
TNFα in sciatica.
Cell cultures have shown TNFα to be a major component of NP (39). In Olmarker's model, TNFα was detected in NP cell cultures by immunohistochemistry with use of a porcine anti-TNFα mAb (39), and in a rat model of disc herniation, TNFα was identified in NP cells by use of a rat anti-TNFα antibody (49). Chronic compression of the rat sciatic nerve demonstrated an increased number of TNFα-positive cells in Schwann and endothelial cells (50). Schwann cells were shown to produce TNFα in vivo, in particular during the first week after radicular insult (50, 51). Immunohistochemical analysis revealed that endoneural activity of endogenous TNFα in a rat CCI model was accentuated only during the first few days after compression (52). In an experiment involving crush injury of rat sciatic nerve, local production of TNFα doubled after 12 hours, remained elevated 3 days after the insult, and returned to baseline levels 14 days after injury (53).
TNFα in animal models.
Treatment of rat nerve roots with exogenous TNFα produced significantly greater neuropathologic damage and behavioral disorders compared with treatment with saline solution (54). In particular, these abnormalities were similar to those observed with NP treatment (24). These effects were confirmed by endoneural injection of TNFα into the sciatic nerves of rats, causing painful neuropathy and histologic changes (32). In an experiment involving Olmarker's model, in 29 pigs treated with NP, retroperitoneal fat, interferon-γ (IFNγ), IL-1β, or TNFα, only TNFα caused changes in NCV on day 7; these changes were similar to those produced by application of NP (Table 3) (55). Treatment of normal rat dorsal root ganglia (DRG) with TNFα provoked persistent allodynia beyond the treatment duration, which was more marked and more prolonged in cases of prior nerve compression compared with that provoked by compression alone (43).
Table 3. Evolution of nerve conduction velocity in pigs, 7 days after application of fat, nucleus pulposus, tumor necrosis factor α, interferon-γ, or interleukin-1β*
Although treatment with NP or TNFα in animal models decreased NCV, to extrapolate this finding to humans, one must demonstrate that TNFα can induce radicular pain. TNFα provoked a painful process and hyperalgesia in animal models of local inflammation or peripheral neuropathy (46). In the experiments with rats conducted by Olmarker et al, videorecordings demonstrated that disc incision or displacement of the DRG of L4 did not change animal behavior, but a combination of mechanical and chemical treatment produced changes in hyperalgesia-linked behavior, which disappeared on day 14 (20). The same model was used in 32 rats, 22 of which were subjected to incision at L4–L5 and displacement of the DRG of L4; 12 of the rats then received intraperitoneal injections of infliximab, a chimeric anti-TNF mAb, for 7 days, and 10 rats were untreated. Untreated rats showed changes in hyperalgesia-related behavior, which was reduced in the 12 rats treated with infliximab (56). Thus, TNFα had a role in pain-related behavioral changes in these animals, particularly when combined with mechanical injury.
Anti-TNFα therapy in animal models.
Thalidomide, which reduces the production of TNFα by macrophages, attenuated thermal hyperalgesia and mechanical allodynia in the rat CCI model (46). The use of a polyclonal anti-TNFα antibody in the murine CCI model had beneficial effects on thermal hyperalgesia and mechanical allodynia (57). The effects of anti-TNFα polyclonal antibody, IL-1 receptor, or a combination in a mouse CCI model showed that the combination had a more marked additive effect on thermal hyperalgesia and mechanical allodynia (58). Etanercept (soluble receptor of TNFα) administered locally or systemically had a more beneficial effect on thermal hyperalgesia and mechanical allodynia in a murine CCI model of the sciatic nerve than did human immunoglobulins (59), and local treatment with 87.5 μg of etanercept had a greater effect than systemic administration of 100 μg of etanercept. In a study investigating whether TNFα caused reduced NCV and formation of intraneural thrombi induced by NP treatment, NCV was restored with intravenous injections of 100 mg infliximab or subcutaneous injections of 12.5 mg etanercept (60) (Table 4), and changes in nerve fibers and the formation of intraneural thrombi were minimal with such treatment.
Table 4. Evolution of nerve conduction velocity in 60 pigs, 7 days after application of nucleus pulposus plus either infliximab, etanercept, heparin, or saline solution*
Nerve conduction velocity, msec
Values are the mean ± SD. Values from the control group in earlier experiments by Olmarker et al were 76 ± 11 msec (see ref.12).
Local treatment with anti-TNFα mAb reduced the response in a rat model of NP-induced nociceptive response of neurons of the dorsal horn, particularly after nociceptive stimulation (61). In this model, NP treatment increased the level of rat brain-derived neutrophic factor (BDNF), a growth factor modulator of nociceptive information, in the spinal cord and ganglion of the dorsal root (62) and the number of ganglion neurons in the dorsal root that were immunoreactive to BDNF, an effect inhibited by infliximab (63). NP in the rat produced less-marked lesions, as observed histologically after treatment with infliximab, which eliminated TNFα in the DRG (64). Intraperitoneal injection of infliximab prevented pain-related behavioral changes, such as mechanical and thermal hyperalgesia, in a rat model of combined disc incision and displacement of the adjacent spinal nerve (65).
Summary of animal experiments on TNFα in sciatica.
Animal experiments have thus revealed some part of the role of TNFα in sciatica, i.e., involvement in the pathophysiology of nerve dysfunction, sensitizing roots previously exposed to mechanical deformation, and discovery in the NP and Schwann cells. Similar to effects observed with NP treatment, TNFα causes electrophysiologic, histologic, and behavioral abnormalities that appear to be enhanced under mechanical compression. Local production of endogenous TNFα occurs at an early stage of the disease process. TNFα-blocking drugs reduce or inhibit abnormalities induced by NP treatment, an effect that varies in intensity according to the timing of the intervention in relation to initiation of the disease process.
Clinical studies in humans.
With experimental findings indicating a central role of TNFα in the pathogenesis of disc herniation–induced sciatica, TNFα blockers have been evaluated in humans and have shown dramatic efficacy in noncontrolled studies. In an open-label study, Karppinen et al evaluated the effects of a single infusion of infliximab (3 mg/kg) in 10 patients with disc herniation–induced severe sciatica (mean duration of pain 7.2 weeks, range 2–12 weeks) (66). The initial mean ± SD radicular pain score (78.7 ± 18.7 mm on a 0–100-mm visual analog scale [VAS]) was reduced by 49% 1 hour after infusion; after 1 week, the pain score was 26.0 ± 21.2 mm, and at 3 months, it was 5.2 ± 6.6 mm (n = 5 patients). Similar results were observed with the evolution of low back pain, straight leg raising, lumbar flexion, and the Oswestry Disability Index score (67). The 10 patients returned to work at 4 weeks, and no patient required surgery or experienced side effects of infliximab treatment.
One open-label study evaluated the efficacy of etanercept (25-mg subcutaneous injection on days 1, 4, and 7) in 10 patients with sciatica (mean ± SD duration 2.7 ± 1.8 weeks, radicular pain score >50 mm on a 100-mm VAS) (68). The intensity of the radicular and low back pain and functional handicap as measured by scores on the Oswestry Disability Index and the modified Roland-Morris Disability Questionnaire (69) improved significantly on day 10 after treatment in all 10 patients, and 9 patients continued to experience improvement until week 6 (Table 5).
Table 5. Evolution of sciatica in 10 patients treated with etanercept*
Values are the mean ± SD. VAS = visual analog scale. See ref.67.
Experimental findings not supporting the involvement of TNFα and the use of TNFα blockers in humans
Identification of TNFα in human intervertebral disc.
Although some studies have shown in cell cultures that TNFα is a component of NP, others have demonstrated barely detectable TNFα in culture media of herniated discs and no secretion at all by disc-derived cells stimulated by lipopolysaccharide in vitro (n = 16); only half of the studies of human herniated discs detected TNFα. Moreover, no studies have shown that the level of TNFα is higher in herniated discs from patients with radiculopathy compared with control patients (i.e., patients with degenerative disc disease without radiculopathy). Indeed, Burke et al compared the level of TNFα, IL-1-β, IL-6, IL-8, and PGE2 in disc tissue from patients undergoing surgery for sciatica with that in tissue from patients undergoing fusion for discogenic low back pain. Those investigators found no specimens producing TNFα or IL-1β but, rather, significant quantities of IL-6, IL-8, and PGE2 were produced in both groups (70). Weiler at al reported a positive correlation between TNFα and intervertebral disc degeneration; however, despite the study demonstrating increased TNFα immunopositivity in surgical samples compared with autopsy samples, the surgical samples were derived from a mixture of both herniated and degenerated intervertebral discs, and thus it was unclear whether increased TNFα immunopositivity was observed in both disorders or only in the setting of herniation (71).
Several substances other than TNFα may explain the effects of NP. The effect of doxycycline in inhibiting NP activity in Olmarker's model is interesting in that doxycycline inhibits not only TNFα but also IL-1 (72), IFNγ, NO synthase (73), and metalloproteinases (38, 72). These substances, particularly IL-1 and IFNγ, act in synergy with TNFα, have neurotoxic potential (74), and are also inhibited by corticosteroids and cyclosporine (33). Larsson et al showed that treatment with TNFα or IL-1β reduced axonal outgrowth from cultured rat DRG, thus suggesting neurotoxic potential, but not as pronounced as with NP (75); the 2 cytokines may act synergistically, both together and with other substances (76), but TNFα and IL-1β were not tested together in this study.
Short-term effect of TNFα blockers in animal models.
Several animal experiments have shown that local production of endogenous TNFα occurs at an early stage of the disease process and is short lived (50–53). Other studies showed that the effect of TNFα antibodies depended on the timing of administration.
Homma et al (45) reported that focal application of soluble TNFα receptor on DRG reduced allodynia in the model of chronic DRG compression, whereas the effect continued only for the first 3 postoperative days during a 7-day application period. Schäfers et al (77) reported that treatment with etanercept that was started 2 days before spinal nerve ligation attenuated allodynia, whereas treatment that was started 1 day or 7 days after spinal nerve ligation had no effect. Rat anti-TNFα mAb reduced allodynia caused by epidural NP treatment in rats, when this treatment was administered immediately after or 6 days after surgery; however, late administration (20 days after surgery) had no antiallodynic effect (78). Selective injury of rat spinal motor fibers by transecting the L5 vertebral root, to induce persistent mechanical allodynia and thermal hyperalgesia in bilateral hind paws, up-regulated TNFα and TNFRI in both DRG and the spinal cord, beginning 1 day and lasting ∼2 weeks after surgery (79). Thalidomide, an inhibitor of TNFα synthesis, applied early (2 hours before surgery), effectively alleviated both the abnormal pain behaviors and the up-regulated TNFα, but treatment 7 days after surgery had no effect on established mechanical allodynia and thermal hyperalgesia. Thus, TNFα may play an important role in the initiation but not maintenance of the neuropathic pain produced by selective motor fiber injury.
These findings suggest that TNFα plays an important role only in the early stage of allodynia, and therefore, TNFα blockade may have an antiallodynic effect only during this critical period. In the clinical situation, administration of TNFα antibody early after the onset of sciatica may thus have a greater effect on pain relief; however, the administration of TNFα blockers in the setting of induced lesions, as with animal models, is impossible in humans, thus challenging the efficacy of anti-TNFα therapy in disc herniation–induced sciatica in humans.
Lack of evidence from randomized controlled trials.
The results of the first randomized, controlled, double-blind trial comparing infliximab and placebo for sciatica were inconclusive (80). That trial included 40 patients who had MRI-verified disc herniation–associated sciatica and symptoms of radicular pain (mean ± SD duration of symptoms 7 ± 3 weeks, range 2–12 weeks) and neural entrapment (straight leg raising ≤60°). Infliximab (a single infusion of 5 mg/kg; n = 21 patients) had no greater effect than placebo (saline solution; n = 19 patients) for pain symptoms or functional handicap, number of days on sick leave, and number of discectomies. The long-term results (1 year) were similar, and 3 nonserious adverse events (rhinitis, diarrhea, and otitis media with sinusitis maxillaris) were experienced by patients in the infliximab group (81). Moreover, MRIs obtained from 21 patients (11 receiving infliximab and 10 receiving placebo) showed that infliximab did not appear to interfere with resorption of disc herniation over 6 months (82).
A phase II, multicenter, randomized, double-blind, placebo-controlled clinical trial of 3 weeks' duration evaluated treatment with oral REN-1654, 100 mg daily, in 74 patients with lumbar radicular pain (83). REN-1654 belongs to a chemical class of compounds known as benzamides and has been shown to act as a functional TNFα antagonist. The group that received REN-1654 showed a trend toward overall pain relief, but the primary efficacy end point (improvement in daily spontaneous leg pain after 3 weeks) was not significantly different between the 2 groups.
Safety of anti-TNF therapy.
We have to take into account possible adverse effects generated by anti-TNF therapy. Indeed, a meta-analysis assessed the harmful events (malignant disease and serious infections) occurring in 9 randomized, placebo-controlled trials of the 2 licensed anti-TNF antibodies (infliximab and adalimumab) used for ≥12 weeks in patients with rheumatoid arthritis (84). The pooled odds ratio for malignant disease was 3.3 (95% confidence interval [95% CI] 1.2–9.1), with a dose-dependent increased risk, and that for serious infection was 2.0 (95% CI 1.3–3.1). Although these results have been much debated (85, 86), several studies have shown the relative risk of serious infection in patients treated with TNF blockers to be 2 or 3 (87, 88). However, in the case of sciatica, anti-TNF therapy is short lived (i.e., a single infusion of infliximab or 3 injections of etanercept), and the risk might be lower, but this possibility needs confirmation.
Results from animal experiments solidly establish a key role for TNFα in the pathophysiology of sciatica; however, several substances other than TNFα may explain the effects of NP in these experiments, and TNFα might be only one of the pieces in the puzzle. TNFα seems to have a role only in the initial stage of sciatica, thus challenging its use as treatment for disc herniation–induced sciatica in humans. Available results from controlled studies with anti-TNF–blocking agents in humans are disappointing, and evidence from current controlled studies are therefore eagerly awaited.
Dr. Goupille had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.