Tuberculosis of the Spinal Cord

Abstract

Tuberculosis involving the spinal cord is associated with high mortality and disabling long-term sequelae. Although tuberculous radiculomyelitis is the most frequent complication, pleomorphic clinical manifestations exist. Diagnosis can be challenging among patients with isolated spinal cord tuberculosis due to diverse clinical and radiological presentations. The principles of management of tuberculosis of the spinal cord are primarily derived from, and dependent upon, trials on tuberculous meningitis (TBM). Although facilitating mycobacterial killing and controlling host inflammatory response within the nervous system remain the primary objectives, several unique features require attention. The paradoxical worsening is more frequent, often with devastating outcomes. The role of anti-inflammatory agents such as steroids in adhesive tuberculous radiculomyelitis remains unclear. Surgical interventions may benefit a small proportion of patients with spinal cord tuberculosis. Currently, the evidence base in the management of spinal cord tuberculosis is limited to uncontrolled small-scale data. Despite the gargantuan burden of tuberculosis, particularly in lower and middle-income countries, large-scale cohesive data are surprisingly sparse. In this review, we highlight the varied clinical and radiological presentations, performance of various diagnostic modalities, summarize data on the efficacy of treatment options, and propose a way forward to improve outcomes in these patients.

INTRODUCTION

Spinal cord involvement in tuberculosis (TB) is a devastating yet under-recognized entity.^[1] Nearly half of the patients with tuberculous meningitis (TBM) may concomitantly harbor spinal cord and spinal nerve root pathology.^[2] Secondary spinal cord involvement may also occur due to vertebral TB.^[3,4] Although the latter is often discussed in the literature, there exists a paucity of information on primary spinal cord TB (SCTB). Identification of spinal cord involvement in TBM is imperative as it may lead to portentous complications including adhesive radiculomyelitis, abscess, tuberculomas, and syrinx formation.

Despite the widespread TB burden in low- and middle-income countries, large-scale data on tuberculous spinal cord involvement are lacking. Most descriptions of spinal TB are focused on the osseous component and its complications, and information on non-osseous spinal TB emanates from small single-institution cohorts, case series, and case reports. SCTB poses specific challenges toward clinical recognition, diagnosis, and therapy. Diagnostic flow heavily depends on radiological features. The optimum duration of anti-tubercular therapy (ATT) for meningeal TB or spinal TB is variable in different guidelines, ranging from 9–12 up to 18 months.^[5–7] Even after ATT, radiculomyelitis, and syrinx formation may complicate the post-treatment course. Apart from ATT, adjunctive strategies such as immunomodulation are routinely used in severe forms of TB, such as TB meningitis. The impact of such interventions on SCTB is unclear. Moreover, fibrinolysis has been employed with variable benefits in small series of patients with adhesive arachnoiditis.

In this review, we examine the pathophysiology, diagnostic and therapeutic challenges, and outcomes in spinal tuberculous meningitis, with the aim of synthesizing current pertinent literature focused on non-osseous manifestations and overcoming specific gaps in the literature with respect to diagnosis and treatment.

SEARCH METHODOLOGY

We searched MEDLINE (PubMed) from 1953 to December 31, 2021. We identified studies on epidemiology, pathogenesis, clinical features, diagnosis, treatment, and outcomes in SCTB. The exact search terms used are provided in the supplementary appendix. We mainly selected articles from the past 10 years but also referenced important older reports and seminal descriptions. Articles in English were included. We also cross-searched reference lists of articles identified by this search strategy. We included systematic reviews and meta-analyses, randomized trials, case series and case reports, and animal research, wherever relevant.

PATHOGENESIS AND PATHOPHYSIOLOGY

Tubercular radiculomyelopathy (TBRM) comprises all atypical forms of spinal tuberculosis such as tubercular myelitis, intradural tuberculosis, and spinal cord complications of tubercular meningitis.^[8]Mycobacterium tuberculosis (MTB) gains entry to the host through droplet inhalation. Bacterial phagocytosis by macrophages triggers a cascade of inflammation resulting in the primary complex. During the short period of bacteremia, MTB can spread hematogenously to the central nervous system (CNS) and form granulomas on meninges, sub-pial, and sub-ependymal surfaces known as “Rich foci,” which may later rupture or grow.

The first description of spinal arachnoiditis leading to subarachnoid obstruction and myelopathy was given by Sir Victor Horsley.^[9] In 1947, Ransome and Montiero published four cases of tuberculous myelopathy in the absence of Potts’s disease, contrary to the belief that TBRM was always a complication of vertebral tuberculosis.^[10] Dastur reviewed 74 cases of tuberculous paraplegia without evidence of Pott’s spine.^[11] He observed extradural granulomas and arachnoid lesions without dural involvement in 64% and 20% of patients, respectively. Intramedullary lesions and subdural extramedullary lesions were present in 8% each.

SCTB originates in one of three ways: 1) rupture or evolution of hematogenously disseminated tubercular foci on the spinal meninges or spinal cord, 2) downward extension of intracranial exudates to subarachnoid space producing spinal arachnoiditis, and 3) extension of adjacent vertebral disease to spinal meninges.^[1,12,13] The commonest mechanism is via the hematogenous route.

Spinal leptomeningitis progresses over three phases: 1) an initial radiculitis phase characterized by nerve root inflammation, 2) the arachnoiditis phase defined by ongoing fibroblast proliferation and collagen deposition causing nerve root adhesions, and 3) the final adhesive arachnoiditis phase identified by atrophied nerve roots encapsulated in dense collagen.^[14] The subarachnoid space in patients with spinal TBM contains thick gelatinous exudates, encasing the spinal cord and radicles partially or completely. There may be hyperemia and edema of the cord and radicles because of inflammation.^[8,13,15] The exudates may cover several spinal segments and are predominantly located posteriorly, possibly due to the prolonged supine position of patients.

The pathogenesis of TBRM shares similarities with models of experimental allergic encephalomyelitis (EAE).^[16] The characteristic features of TB myelitis include inflammatory cell infiltration, perivascular demyelination, axonal damage, and activation of glial tissue [Figure 1].^[12,13] Inflammatory exudates can damage the parenchyma by direct infiltration or cause subarachnoid block due to elevated cerebrospinal fluid (CSF) protein. Occasionally, extensive exudates compress the cord without direct invasion. Demyelination of the entrapped nerve roots and white matter damage leads to axonal loss. Occlusion of the spinal vessels by necrotizing granuloma or vasculitis can cause spinal cord infarction.^[13,16]

Figure 1: Pathological changes associated with early and late tuberculous radiculomyelitis

In the latter stages of the disease, the tenacious exudates get organized, leading to the clumping of fibrin-glazed roots. There may be atrophy of nerve roots as well, together known as chronic adhesive arachnoiditis. Arachnoid adhesions obstruct CSF flow with the formation of loculations.^[8,11] Irregular subarachnoid space can cause occlusion of vessels leading to damage to the cord and radicles; cord atrophy is seen in the final stages.

Syrinx develops as a late complication of TBM, though the precise mechanism of syrinx formation is unclear. One proposed mechanism is obliterative endarteritis causing ischemic injury and softening of the spinal cord, resulting in cavitation. Another hypothesis is that of obstruction of CSF flow due to local scarring, forcing the CSF into the central canal through Virchow–Robin (VR) spaces; the dilated VR spaces coalesce to form a syrinx.^[17,18] Savoiardo proposed an alternate hypothesis of syrinx formation. Neck movement-induced elongation of the spinal cord (fixed by adhesions) results in the squeezing of the necrotic pulp or cystic spaces in the cord.^[19] The disruption of the spinal cord at areas of least resistance causes upward extension of the syrinx; on rare occasions, the syrinx can communicate with the subarachnoid space.^[20]

Similar to TBM, TBRM might be a delayed hypersensitivity reaction to tubercle bacilli protein. TBRM commonly develops during the treatment of TBM. The likely explanation for this paradox is the recovery of the delayed hypersensitivity response with treatment. Enhanced immune reaction to tubercular antigen, released during ATT due to mycobacterial death, is a putative mechanism.^[21]

To summarize, TB affects the spinal cord by four mechanisms: 1) obstruction of venous drainage associated with meningitis, causing edema of the border zone, 2) vasculitis-related or thrombotic occlusion leading to ischemic myelomalacia, 3) development of intramedullary tuberculomas with caseation necrosis, and 4) rarely, vascular occlusion resulting in cord infarction.

CLINICAL PRESENTATION

Tuberculous involvement of the spinal cord is highly variable in presentation, ranging from exclusively lower motor neuron (LMN) or upper motor neuron (UMN) to mixed UMN/LMN type of neurological involvement [Figure 2]. In a case series from northern India amongst 71 patients with TBM, 33 (46.4%) had clinical features of the spinal cord and nerve root involvement. Paraparesis was variably of UMN (6 patients; 8.4%), LMN (10; 14%), and mixed (6; 8.4%) type.^[2]

Figure 2: Clinical phenotypes in spinal cord tuberculosis

In a retrospective study from South Africa, 274 patients with spinal tuberculosis without bony involvement were assessed amongst both HIV-infected and HIV-uninfected individuals. The majority (76%) of patients were HIV-infected.^[22] Spinal cord involvement frequently followed TBM as a paradoxical reaction. Paradoxical reaction presenting as spinal involvement was seen in 24% of HIV and 14% of HIV-uninfected patients. Radiculomyelitis was observed in 210 (77%) patients. Other common manifestations included spondylitis (39%), subdural abscess (15%), and intramedullary tuberculoma (12%).

Radiculomyelitis

TBRM is the most frequent manifestation of tuberculous spinal cord involvement. TBRM is a broad term that is variably employed to include tubercular myelitis, arachnoiditis, spinal cord tuberculoma, spinal cord edema, or infarction.^[21] It is this variation in descriptive case reports that lends challenges to estimates of its exact incidence. Inflammation of the arachnoid membrane, called arachnoiditis, leads to the involvement of nerve radicles encased by the arachnoid membrane as they traverse out of the spinal canal. Clinically, arachnoiditis presents as an LMN syndrome affecting the lower limbs typically, with hypotonia, areflexia, radicular pain, paresthesia, and in long-standing situations, muscle atrophy. Bladder and bowel involvement in the LMN pattern may also occur. Lumbosacral radicles are most commonly involved, giving rise to cauda equina syndrome. Frequently, conus syndrome may also occur. This is clinically characterized by prominent sphincter disturbances, perianal or saddle anesthesia, and extensor plantar response with absent ankle reflexes, representing an admixture of UMN and LMN features. TBRM may appear at variable periods, even in patients who are adequately treated.^[21,23] It has also been described to develop as a paradoxical response during therapy.^[24] Multifocal loculations and adhesive arachnoiditis are contributors to syrinx development, treatment non-response, and poor outcome.^[25]

Tubercular spinal abscess

Intramedullary spinal cord abscess is an exceedingly rare presentation.^[26–28] Typical presentation is myelopathy, with paraparesis of acute or subacute onset, sensory deficits, and bowel/bladder involvement. The thoracic spinal cord is the most affected, followed by the cervical and lumbar regions. This predilection is determined by differential blood flow to the spinal cord, with thoracic regions receiving nearly 45% of total cord blood flow. Diagnosis may be challenging as, unlike the brain, ring enhancement on MRI may be minimal to absent in the spinal cord.^[29] Epidural and spinal subdural tubercular abscesses may also occur, rarely without osseous involvement.^[30,31]

Syrinx formation

Syringomyelia is an uncommon late complication of TBM although both early and paradoxical development on ATT has been reported.^[32] The development of syrinx following TBM may be delayed by several years of symptom-free intervals. Early syrinx formation was first reported by Daif et al. in two cases, developing after 11 days and 6 weeks of the onset of tubercular meningitis.^[33] In a series of 10 HIV-infected patients with post-tubercular syrinx formation, syringomyelia developed after a median of 21 months after initial diagnosis.^[34] In contrast, in the same report, the authors identified (via literature review) that the duration among 46% of HIV-uninfected patients for post-tubercular syrinx development was longer (>4 years), compared to 8% of HIV-infected individuals.

Spinal cord infarction

Obliterative endarteritis or infective thrombosis of the spinal vessels is a potential complication. However, spinal cord strokes have been infrequently described. In one patient, paraparesis developed after ATT for TBM, which was attributed to spinal infarction, demonstrated on diffusion-weighted imaging.^[35] Reports of spinal cord infarction in the setting of tubercular meningitis are scarce. This is surprising, considering the frequent occurrence of infarcts in the brain. Up to 41% of patients with CNSTB display infarcts on imaging.^[36] Whether this discrepancy is due to the under-recognition of spinal infarcts, or whether spinal vessels are less susceptible to obliterative endarteritis, is unclear.

Tuberculomas

Spinal tuberculomas, which may occur in the intramedullary, extramedullary intradural, or extradural location are rare. It is estimated that intramedullary spinal tuberculomas occur in 2/100,000 cases of TB and 2/1,000 cases of CNS TB.^[37] There are above 100 individual case reports and a few small case series indexed in PubMed that report intramedullary spinal tuberculoma. The usual location is the thoracic spinal cord; however, these have been reported from other regions, including the cervical cord and conus. Literature on extramedullary intradural tuberculomas is even scarcer. These may mimic en plaque meningioma.^[38] TBM may concur with spinal cord tuberculoma, but more commonly, the latter follows the former.^[39] Tuberculomas of the spinal cord have been shown to respond well to ATT, with or without concomitant steroids.

Isolated myelitis

Tuberculosis is a rare cause of transverse myelitis (TM). Both short-segment and rarely, longitudinally extensive transverse myelitis (LETM) have been reported in the literature and should enter the differential diagnosis for TM in countries in which tuberculosis is endemic. In a review of 10 patients with tuberculous LETM, extensive myelitis extending up to the conus was described.^[40] Most of these patients presented with acute or subacute onset paraplegia with bladder involvement, in association with fever, headache, or altered sensorium. The duration from symptom onset to presentation ranged from 1 day to up to 2 months. Of six patients tested for aquaporin-4 antibodies, all were negative.

Cervicodorsal involvement is the most frequent.^[41] Several case reports have described LETM in tuberculosis and reported an association with neuromyelitis optica spectrum disorders (NMOSD).^[42,43] However, the current evidence is tenuous and needs to be corroborated in larger case-control studies. The underlying pathogenesis is likely to be immune-mediated. Most patients demonstrate excellent clinical and radiological responses to a combination of steroids and ATT.

Vertebral tuberculosis with cord involvement

Osseous vertebral involvement may lead secondarily to spinal cord involvement. The dorsal vertebral region has increased predilection for this complication, considering the narrow width of the spinal canal in this region, in combination with the physiological kyphosis at the thoracic column, which physically may push the tuberculous tissue toward the cord.

Paraplegia in spinal osseous tuberculosis may be early-onset (paraplegia of active disease) or late-onset (paraplegia of healed disease), as per Hodgson’s classification.^[44] Various mechanisms underlying early-onset paraplegia include mechanical factors (compression via granulation tissue/abscess, vertebral collapse/instability, gibbus formation), apart from arachnoiditis, spinal cord infarction, myelitis, and tuberculomas described above. Paraplegia of healed disease may be due to bony factors such as severe kyphosis and pachymeningitis.

Impact of HIV on presentation

Southeast Asia and Africa are endemic for TB-HIV co-infection. Although the effects of this co-infection are well studied in pulmonary TB, there is a paucity of data regarding the clinical manifestations and outcomes for extra-pulmonary TB, and lesser for spinal cord involvement. In non-HIV patients, TBM with concomitant spinal meningitis occurs in around 10% of the patients.^[45] This association is, however, stronger in patients with HIV, with one report documenting this figure to be around 48%.^[22] TB-HIV immune reconstitution inflammatory syndrome (IRIS) in the spinal cord predominantly presents as radiculomyelitis^[46,47] though epidural abscess and spinal tuberculomas have been documented as well.^[1]

Modi et al. found that of 97 patients presenting with non-traumatic myelopathy in South Africa, 50% had HIV, of whom 50% had concomitant tuberculosis.^[48] Bhigjee et al. similarly found tuberculosis to be the most common cause of myelopathy among people living with HIV.^[49] In another similar large series from South Africa comprising 216 patients, acute-onset myelopathy/cauda equina syndrome in HIV-positive patients was largely attributable to tuberculosis, with non-spondylitis forms being more common than spondylitis at a lower CD4 count.^[50] Marais et al. observed that syringomyelia was a relatively early manifestation (within 4 years) in HIV patients with neurological TB compared to non-HIV patients, with relatively high mortality and poor outcomes.^[22] In another large case series from South Africa involving 274 patients with spinal TB disease without bony involvement on plain X-ray, 76% of the patients were co-infected with HIV. The clinical presentation and outcomes were similar in both HIV and non-HIV cohorts. However, compared to non-HIV patients, patients with HIV more commonly presented with radiculomyelitis had a significantly higher incidence of paradoxical spinal cord disease, lower rates of vertebral destruction in patients with concomitant spondylitis, higher incidence of TBM, and a lower CSF glucose.^[22]

Paradoxical worsening in spinal cord tuberculosis

Paradoxical worsening, characterized by the worsening of preexisting lesions or the development of new lesions, complicates almost a third of patients with SCTB. The paradoxical worsening is important as it may confuse the clinicians and elicit doubts regarding the diagnosis and possible drug resistance. Younger age, HIV infection, a large burden of disease, mycobacterial smear positivity, and elevated proteins are risk factors for paradoxical worsening. Many unique features are associated with SCTB. Although most paradoxical worsening in TBM is not of significant long-term clinical consequence, residual sequelae are common among patients with SCTB. Large vessel infarcts characterizing TBM are uncommonly recorded in autopsy and radiological studies of SCTB. Imaging and CSF findings may not correlate with clinical worsening in the spinal cord.^[51]

Three patterns of paradoxical worsening occur in patients with spinal TB: 1. new onset spinal cord involvement in patients with TB elsewhere, 2. worsening of existing spinal cord lesions, and 3. remote worsening, for example, development of syrinx. Patients with a large burden of diseases, such as military TB, often present with new onset brain or spinal cord involvement. The release of mycobacterial antigens on antimycobacterial therapy is the putative mechanism. Almost 50% of patients with SCTB worsen while on treatment. Preliminary data suggests early worsening (less than 4 weeks of antimycobacterial therapy) may be associated with younger age, female sex, and associated with CSF changes. Late worsening, while on therapy, is usually secondary to worsening tuberculomas within or on the surface of the cord. Remote events such as syrinx formation rarely complicate SCTB even after the completion of antimycobacterial therapy.

EVALUATION

History

SCTB generally occurs secondary to the spondylitic form of tuberculosis, after rupturing of granuloma into the spinal cord, or due to secondary seeding from TBM. The patient may present with constitutional symptoms such as fever, night sweats, weight loss, and loss of appetite. The patient may complain of symptoms of radiculomyelitis, spinal meningitis, bowel/bladder involvement, transverse myelitis, tuberculous abscess, and syrinx formation. Headache, fever, vomiting, and altered sensorium may suggest concurrent CNS involvement, which warrants urgent attention. Other possible clues may include hemoptysis, cough, breathlessness, lymphadenopathy, and recurrent urinary tract infections. History of earlier TB infection, known or possible TB exposure, and residence in endemic regions must be elicited to further consolidate this diagnosis.

CSF findings

CSF findings in patients with spinal cord disease are similar to those seen in patients with TBM, with predominantly lymphocytic leukocytosis (10–100 cells/mL), high protein, and low CSF glucose [Table 1]. Patients with concomitant HIV infection and those with early meningeal disease may have predominant neutrophilic leukocytosis.^[52] In patients with subarachnoid block, CSF protein concentration may be as high as 2 g/dL. CSF smear for AFB, TB cultures, and newer molecular tests may help clinch the diagnosis. Isolated forms of spinal tuberculosis such as arachnoiditis in the absence of meningitis, may be more difficult to diagnose. A targeted biopsy followed by microbiological, molecular, and histopathological analyses may help in these cases.

Table 1: Cerebrospinal fluid findings in various forms of spinal cord tuberculosis^{[ 22 ]}

MOLECULAR TESTS

Xpert MTB/Rif

Xpert MTB/Rif is a semi-nested cartridge-based polymerase chain reaction (PCR), which provides information regarding the presence of TB and rifampicin resistance within 2 h of the sample being processed using the wild-type mycobacterial DNA through five partially overlapping fluorescent probes. Limited diagnostic sensitivity (55–80%), difficult interpretation of rifampicin resistance in cases of “very low load,” and the availability of Xpert Ultra preclude its use in the diagnosis of TBM and SCTB.^[53]

Xpert Ultra

Xpert MTB/Rif Ultra, a cartridge-based nested PCR test, is now endorsed by the World Health Organization (WHO) for the diagnosis of TBM.^[54] Recently, two large prospective studies carried out in the African cohorts found that the sensitivity of CSF Xpert Ultra was significantly higher than Xpert MTB/Rif and TBMGIT in the diagnosis of TBM in the HIV population.^[55,56] A Vietnamese study,^[57] however, failed to find a significant difference between CSF Xpert Ultra and CSF Xpert MTB/Rif to diagnose TBM, which may be explained by the fact that only 15% of the patients in the Vietnamese cohort were HIV positive, who are expected to have higher bacillary loads in the CSF.

Pyrosequencing

Pyrosequencing (PSQ) is real-time PCR providing information on extremely drug-resistant (XDR)-defining mutations within 6 h in the MTB genome. A small pilot study comprising 13 patients with suspected TBM found that pyrosequencing, when performed directly on the CSF samples, established the diagnosis in 84.6% of the patients compared to Xpert MTB/Rif (15%) and TBMGIT (30.7%).^[58] It detected six patients with drug-resistant TB, who were not detected by TBMGIT culture. A further larger study conducted on 100 patients with suspected TBM found that the diagnostic accuracy of pyrosequencing was significantly higher compared to Xpert MTB/Rif and TBMGIT (98% vs. 43% vs. 45%) in the diagnosis of TBM.^[58] Despite the limitations of the small sample size, the retrospective nature of the study, lack of information about HIV status, and the inherent limitations of the test (expensive, requirement of molecular expertise), the ability of this test to provide information on XDR-defining mutations within 6 h when performed directly on the CSF sample with excellent diagnostic performance may have significant implications on clinical outcomes, especially in regions endemic for drug-resistant TB.

Next-generation sequencing

Newer molecular techniques such as targeted next-generation sequencing and whole-genome sequencing are presently being evaluated in the diagnosis of tuberculosis. Although pyrosequencing provides information on resistance to isoniazid, rifampicin, fluoroquinolones, and aminoglycosides, whereas targeted next-generation sequencing provides additional information on resistance to pyrazinamide, ethambutol, linezolid, bedaquiline, capreomycin, and clofazimine. A recent study on 40 uncultured sputum samples found that tNGS provided results for 39/40 samples with a significantly faster time than phenotypic MGIT (3 vs. 21 days, P: 0.0068).^[59] The utility of these molecular techniques in TBM and spinal cord involvement remains uncertain and may offer an exciting avenue for research.

Imaging

Magnetic resonance imaging (MRI) is the most sensitive tool for the anatomical assessment of the spinal cord and surrounding nerve roots [Table 2]. In the setting of tuberculous infection, MRI provides an accurate estimate of the extent of the involved structures and associated complications. Imaging abnormalities may present as enhancement along the surface of the cord and thickening and enhancement of the nerve roots, the presence of nodular enhancement along the surface of the cord, edema within the cord, and ring-enhancing lesions within the substance of the cord.^[2,60,61] MRI is also a useful tool to assess epidural collection secondary to adjacent tuberculous spondylodiscitis causing cord/nerve compression [Figures 3 and 4]. Meningitis and arachnoiditis are considered the two most common imaging findings, seen in more than 50% of cases in patients with TBRM.^[2,61] The most important imaging differential for nodular enhancement along the surface of the cord/nerve roots would be drop metastases from an intracranial primary, whereas for intramedullary tuberculoma, it would be focal metastasis. Table 3 enlists the imaging differences between tuberculous and non-tuberculous myelitis, supported by Figures 5a and b.

Table 2: Imaging findings in various forms of spinal tuberculosis

Figure 3: Imaging findings in spinal tuberculosis. Sagittal and axial imaging of the spine (T2-weighted images a, b, e, f, g, i, j and post-contrast T1-weighted c, d, h) showing thickening and nodularity along the cauda equina nerve roots (a and b) associated with post-contrast enhancement (c and d) suggestive of radiculitis. A well-defined oval, partially cystic lesion (e) is seen involving the thoracic cord with edema out of proportion to the lesion suggestive of an evolving abscess. Image (f) demonstrates a long-segment cord hyperintensity and expansion suggestive of myelitis associated with the syrinx. Multiple T2W hypointense lesions involving the cord with mild surrounding edema (g) suggest the presence of multiple intramedullary tuberculomas, whereas nodular and ring-like enhancement along the surface of the cord suggest meningeal inflammation and tuberculomas (h). Image (i) demonstrates long-segment myelitis, whereas cord compression secondary to tuberculous spondylodiscitis-associated extradural abscess is seen in image (j)

Figure 4: Sagittal imaging of the cervicothoracic spine (b) (T2-weighted images a and c and post-contrast T1-weighted images b and d) demonstrate inflammatory exudates along the dorsal aspect of the thoracic cord (a) causing anterior displacement of the cord with cord hyperintensity (a) and post-contrast enhancement (b). Image (c) demonstrates scalloping of the thoracic cord with resultant cord edema secondary to chronic arachnoiditis. No obvious post-contrast surface enhancement is seen (d); however, there is an enhancing parenchymal granuloma involving the cervical cord

Figure 5: (a) Imaging findings in non-tuberculous myelitis. Long-segment T2W hyperintensity on sagittal imaging of the cord. Image a demonstrates heterogeneous cord hyperintensity associated with swelling seen with melioidosis of the cord. Image b1 shows long-segment cord hyperintensity extending up to the conus. Digital subtraction angiography (image b2) demonstrates an abnormal tortuous vessel along the surface of the cord confirming the diagnosis of a dural arteriovenous fistula. Image c1 shows a long-segment cord hyperintensity in a patient with hyperacute paraplegia raising suspicion for cord infarct. Diffusion restriction is seen on diffusion imaging (c2) confirming the diagnosis. (b) Image a1 shows localized T2W cord hyperintensity with the scalloping of the cord suggestive of arachnoiditis. During surgery, the thecal sac was found to contain multiple cysticercal cysts. Image a2 shows central cord signal changes in the same patient secondary to cysticercal arachnoiditis. Image b demonstrated short-segment cord T2W cord hyperintensity in a patient with disseminated CMV infection. Images c1 and c2 show a solidary cysticercus granuloma with a ring-like post-contrast enhancement. The lesion showed significant regression with steroids. Image d demonstrates smooth enhancement of the cauda equina nerve roots in a patient with acute inflammatory demyelinating polyneuropathy; whereas nodular enhancement is seen (image e) in a patient with leptomeningeal carcinomatosis

Table 3: Imaging differences between tuberculous vs. non-tuberculous myelitis

Predictors of asymptomatic spinal cord involvement in patients with TBM

Patients with TBM, especially with concomitant HIV infection, are at a higher risk of developing cord involvement. A gadolinium-enhanced MR imaging of the spine in these cases may help in detecting occult nodules/granulomas, if present.^[23] A pilot study performed by Srivastava et al. found that 3/16 patients with TBM with no symptoms suggestive of spinal cord involvement had evidence of spinal arachnoiditis on MRI when performed within 1 month of diagnosis of TBM. These patients had higher CSF protein (520 ± 48.5 mg/dL vs. 300 ± 43.7 mg/dL) compared to those without radiculomyelitis.^[62] Wadia et al. noted patients with TBM with no symptoms suggestive of spinal cord involvement and showed the presence of exudates around the spinal cord and nerve roots.^[8] Gupta et al., in a prospective study comprising 71 consecutive patients of newly diagnosed TBM, found that 11% had evidence of spinal meningitis on MRI despite being asymptomatic.^[2] Univariate analysis showed that high CSF protein and a baseline-modified Barthel index < 12 were significant predictors of spinal cord involvement in patients with TBM, though multivariate analysis did not find these factors to be statistically significant.

Patients with osseous forms of spinal tuberculosis comprise yet another important population, which may be at a higher risk of cord involvement. Sae-Jung et al., in a retrospective study comprising 125 patients with spinal tuberculosis, found that cord signal changes and notable Cobbs angle were significant predictors of neurological deterioration.^[63] A similar study from India found that kyphosis >30 degrees, cord edema, and canal encroachment (>50%) were significant predictors of neurological deficit.^[64] Wang et al. found that elderly patients with cervical and lumbar vertebral involvement were more at risk of developing neurological disturbances.^[65]

To summarize, in patients with TBM, high CSF protein is an important predictive marker of asymptomatic spinal cord disease. Concurrent spine screening by gadolinium-enhanced MRI may be helpful in these cases. In those with osseous spinal cord involvement, old age, kyphosis of more than 30 degrees, and cord edema were important predictors of neurological deterioration.

TREATMENT OF SPINAL CORD TB

What do guidelines say?

The established guidelines for the management of CNS TB largely focus on TBM.^{[5,6,66–68]} There are no specific recommendations to treat SCTB. Management principles for TBRM are derived from the treatment guidelines for CNS TB.

The WHO guidelines for the treatment of drug-susceptible CNS TB recommend the administration of ATT in two phases.^[66] The initial 2 months of the intensive phase are followed by a continuation phase of 7 or 10 months. In the intensive phase, the patient receives a combination of four first-line drugs (isoniazid, rifampicin, pyrazinamide, and streptomycin). In the continuation phase, 2-drug (isoniazid and rifampicin) ATT is given. The British Infection Society (BIS) and National Institute for Health and Care Excellence (NICE) guidelines also recommend at least 12 months of ATT for all forms of CNS TB; isoniazid, rifampicin, pyrazinamide, and ethambutol for initial 2 months followed by isoniazid and rifampicin for 10 months.^[68] The NICE guidelines advocate managing tuberculous involvement of the spinal cord as CNS TB. The American Thoracic Society (ATS) recommends 9–12 months of ATT for meningeal involvement; isoniazid, rifampicin, pyrazinamide, and ethambutol for the initial 2 months followed by isoniazid and rifampicin for 7–10 months.^[6] Similarly, the Index-TB guidelines in India recommend 2 months of isoniazid, rifampicin, pyrazinamide, and ethambutol followed by isoniazid, rifampicin, and ethambutol for 7 months for the treatment of CNS TB.^[5] Whether patients with extensive spinal arachnoiditis will benefit from prolonged ATT is unclear.

Role of ATT in spinal cord TB

TB myelitis

Neurological insufficiency in TB myelitis is secondary to inflammatory activity.^[1] Previous reports suggest an excellent response of TB myelitis to a combination of ATT (isoniazid, rifampicin, pyrazinamide, ethambutol/streptomycin) and adjuvant steroids.^[24,69,70] Early initiation of ATT is important. In two different series, involving four cases of longitudinally extensive transverse myelitis (LETM) because of TB, medical therapy alone resulted in a favorable outcome.^[71,72] In a recent review of 10 reported cases of tubercular LETM, treatment with ATT and steroids resulted in clinical and radiological improvement.^[40] Patients with damaged anterior horn cells have poor recovery with ATT.^[73] The patient should receive a combination ATT for at least 9–12 months. Further extension of therapy in some cases is dictated by an amalgamation of clinical, radiological, and other supportive evaluations that suggest active disease, although guidelines remain unclear on this issue.

Corticosteroids and hyaluronidase may be an adjuvant therapy to ATT. Steroids are often used as a pulse therapy in patients with severe concerns, such as LETM or severe arachnoiditis, or large tuberculomas with significant edema, causing paraparesis, as per the literature. Regimens are variable, with steroid durations ranging from 8 weeks to 6 months.

TB spinal arachnoiditis

Treatment of TB spinal arachnoiditis may be medical or surgical. Medical therapy remains the mainstay of management.^[24,69] As arachnoiditis is diffuse, surgery may be of limited role. If there is a resistant strain, drugs are modified according to the sensitivity pattern.

Intradural extramedullary tuberculomas

Evidence shows that medical therapy alone will be insufficient for the management of spinal intradural extramedullary tuberculoma. ATT combined with surgical excision yields excellent results.^{[24,69,74,75]}

Intramedullary tuberculoma and abscess

Before the MRI era, surgical excision of the intramedullary lesion was the technique of choice for both diagnosis and treatment. Currently, ATT is the preferred treatment for intramedullary tuberculoma and abscess.^[2,21,24,28] Adjunctive corticosteroid therapy results in a favorable outcome. With early diagnosis and treatment, one can avoid complications and unnecessary surgical intervention. Surgery is indicated in the management of spinal intramedullary tuberculoma for 1) a large lesion with rapid progression of neurological deficits, 2) deterioration of neurological status despite medical therapy, 3) inconclusive neuroimaging findings, and 4) paradoxical increase in the size of the lesion after ATT.^[75–77]

Therapy for spinal cord paradoxical worsening

Corticosteroids are the main immunomodulatory therapy in CNS TB. The evidence is largely derived from patients with TBM without HIV infection. Importantly, corticosteroid therapy decreases mortality without impact on disability in TBM. Corticosteroids probably beneficially impact SCTB as well, as close to a third of TBM patients have concomitant spinal cord involvement. However, no randomized trials have evaluated the role of steroids in spinal TB so far. The percentage of paradoxical worsening in SCTB is higher than in TBM. Often paradoxical worsening happens despite receiving adequate steroids. The role of adjunctive immunomodulatory therapy in these individuals is unclear. Adjunctive thalidomide therapy at low doses (3–5 mg/kg/day) produced a favorable response in two children with tuberculous mass lesions involving the spinal cord.^[78] The duration of thalidomide therapy used in these patients was 8 weeks although some advocate shortened regimens of less than a month. However, longer durations have been used in children with optic neuritis (median 2, interquartile range [IQR] 1.3–7.3 months).^[78] Painful paresthesia may develop with longer regimens due to the development of sensory axonal neuropathy. Infliximab, a tumor necrosis factor-alpha inhibitor has been successfully used in three patients with paradoxical worsening involving the brain and spinal cord.^[79] Larger studies evaluating the benefits of steroid therapy and other alternative agents are needed.

Evidence for other interventions

Fibrinolysis

The use of hyaluronidase as an adjuvant in the treatment of tuberculous arachnoiditis has been described in non-randomized studies in the 1980s and 90s [Table 4].^[80–82] In a recent retrospective description of the use of intrathecal hyaluronidase among patients with spinal tubercular arachnoiditis or optico-chiasmatic arachnoiditis, weekly hyaluronidase for 10 weeks was administered, besides ATT and steroids. Of 19 patients with spinal arachnoiditis, 10 were independent at the completion of therapy.^[83] The benefits of fibrinolysis reported from observational literature need augmentation by further evaluation in a randomized trial setting.

Table 4: Studies on the use of fibrinolysis for tuberculous arachnoiditis

Surgical management

Intramedullary tuberculomas, intra-conal abscess

Patients with intramedullary TB can present with a mass-like presentation mimicking intramedullary tumor, sometimes like an intramedullary abscess. The aim of surgery is to perform safe surgical excision of the lesion without causing new neurological deficits. The surgical procedure involves laminectomy done at the level of the lesion. The dura is opened, and CSF is let out to allow the pulsations of the cord. A midline myelotomy is performed after identifying the lesion. After identifying the lesion, a plane is developed to safely excise the lesion. Intraoperative neuromonitoring helps in safe surgical excision. The pia mater is closed by welding technique and the dura is closed primarily.

Syringomyelia

The best management for post-TB syringomyelia is unclear. Often patients are treated by placing a syringe-subarachnoid shunt with mixed results.^[24] A focal laminectomy is performed. The dura and the arachnoid are opened. A small myelotomy is done in the center of the syrinx to open it. One of the shunt tubes is gently guided cranially or caudally into the syrinx. The other end of the tube is placed in the subarachnoid cavity beneath the dentate ligament. This helps to keep the shunt in place. The dura is then closed. Spinal cord function, specifically motor evoked potential (MEP), somatosensory evoked potential (SSEP), and D waves are monitored throughout the procedure to minimize the risk of postoperative neurological deficit.

Outcomes in spinal cord tuberculosis

Outcomes among patients with spinal cord involvement in TB are variable and predictors of outcomes are unclear.

Large studies reporting outcomes of various types of spinal cord involvement are rare. Most available studies are single-center experiences, retrospective in nature, and document outcomes with radiculomyelitis, the commonest presentation. In the available studies [Table 5], poor outcomes were reported in up to 35–85% of the study population despite standard mycobacterial therapy and steroids. This is markedly higher compared to any other form of tuberculosis, including TBM.

Table 5: Outcomes of spinal cord tuberculosis (reporting studies with at least 20 patients; studies chiefly reporting on tuberculous spondylodiscitis were excluded

In one study of 71 patients with TBM, 33 (46.4%) had spinal cord and/or nerve root involvement clinically. MRI demonstrated meningeal enhancement (40; 56.3%), myelitis (16; 22.5%), tuberculoma (4; 5.6%), CSF loculation (4; 5.6%), myelomalacia (3; 4.2%), and syrinx (2; 2.8%). Radiculomyelitis was determined to be associated with high CSF protein (>250 mg/dL) and was associated with unfavorable outcomes (modified Barthel index <12 or death) at 6 months.^[2] In a recent series of 198 patients with tubercular meningitis, 29 (14.6%) had spinal tubercular meningitis, and among those who completed therapy, 84% were dependent for their activities of daily living.^[71]

Unique anatomical location within a restricted space, complex vascularity, and close proximity of neural tracts predispose to poor outcomes. There may be several further reasons for poor outcomes, mainly engendered by the underlying pathology. The development of syrinx/cavitation and myelomalacia leads to permanent sequelae.^[41] Additionally, in patients with tuberculous myelitis, involvement of the anterior horn cells may portend a poor prognosis in terms of motor recovery.^[73] LETM, which frequently spans nearly the entire spinal cord length, is associated with poorer recovery. In addition, we hypothesize that some of these patients are developing spinal cord ischemia, which remains under-recognized, and may portend a poor prognosis.

Future directions

There are many avenues for focused research in spinal cord TB. A discrepancy between what the guidelines enunciate and what neurologists practice at the ground level in India is reflective of the multiple lacunae that exist.^[85]

The most striking concern with SCTB is poor long-term outcomes. At least a third of patients remain severely disabled while the proportion is less than 15% in TBM.^[86] A long-sighted, multi-pronged approach is needed to reduce this disability in SCTB [Figure 6]. Firstly, there is an acute need for the documentation of long-term outcomes with antimycobacterial therapy and steroids in large multicentric cohorts of patients. The pleomorphic presentations and varied anatomical involvements mandate such large studies. Secondly, the mechanism of long-term sequelae in SCTB is unclear. Although the thick characteristic gelatinous exudative inflammation is responsible for adhesive complications such as arachnoiditis, the molecular mechanisms responsible are not elucidated.

Figure 6: Future direction in spinal cord tuberculosis

Large vessel infarctions are uncommon in TB of the spinal cord. Transmural inflammation of the vessels with cellular infiltration, intimal thickening, and subintimal proliferation with or without granulomas are central to pathogenesis for TBM.^[12] Whether similar mechanisms are operational in SCTB needs to be evaluated. Vascular endothelial growth factor (VEGF) and matrix metalloproteinase-9 play a major role in the pathogenesis of TBM.^[87–89] Role of the above mediators in SCTB needs to be studied. Only a proportion of patients with disseminated TB develop adhesive arachnoiditis. Variations encoding a promoter region in the leukotriene A4 hydrolase (LTA4H) gene, a critical link in balancing the pro and anti-inflammatory eicosanoids appear to predict hyper-inflammatory and hypo-inflammatory phenotypes and the subgroup of patients who benefit the most from steroid therapy in TBM.^[90–92] Whether this finding can be extended to SCTB is not clear. The degree of reduction of CSF tryptophan appears to predict survival in TBM. Both mycobacterial factors and host genome single nucleotide polymorphisms appear to play a strong role. Whether tryptophan metabolism plays such a critical role in SCTB is undetermined.^[93,94] Finally, whether adjunctive agents such as tumor necrosis factor alpha-blockers such as infliximab and thalidomide and other immunomodulatory strategies can further improve outcomes of patients who worsen despite antimycobacterial therapy and steroids needs to be studied. The role of aspirin and antifibrinolytics in the management of TBRM also should be studied in randomized trials.

CONCLUSIONS

Diagnosis and management of patients with spinal cord tuberculosis are difficult. Clinical presentations can be diverse and paradoxical worsening frequently occurs. Often, microbiological confirmation and drug sensitivity are lacking, making the diagnosis of resistant TB challenging. Drug resistance may be presumed in a patient who shows suboptimal response or worsening to the first-line regimen, in the absence of drug sensitivity testing. The subsequent regimens should be as per national guidelines for resistant TB.

Currently, the management of SCTB is not different from TBM. Studies evaluating the underlying pathophysiology and efficacy of strategies for augmented mycobacterial killing and controlling host inflammation in SCTB are urgently needed.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.