17 Thoracic Osteomyelitis and Discitis Abstract Vertebral osteomyelitis can result in severe spinal cord compression and spinal deformity. It may present nonspecifically with back pain or with paraparesis in advanced cases. Clinicians should be familiar with the epidemiology, signs, symptoms, and radiographic evidence of thoracic osteomyelitis and discitis to diagnose and treat these conditions expeditiously prior to the appearance of neurological deficits. Treatment consists of antibiotics and potentially surgery, both of which must be tailored to the individual patient and pathogen to optimize outcomes. Keywords: osteomyelitis, discitis, bacterial infection, postinfectious deformity, sepsis Clinical Pearls • Pyogenic vertebral osteomyelitis (PVO) is the most common spine infection in adults. Its insidious onset sometimes delays the clinical diagnosis. • After radiographic diagnosis by computed tomography or magnetic resonance imaging, a biopsy should be performed for tissue diagnosis to guide antibiotic therapy. • Surgery should be considered for diagnosis, debridement of extensive bony destruction, and decompression of the thecal sac in cases of neurological compromise, instability, and deformity. • Although previously controversial, instrumentation in the setting of active infection, such as PVO, has not been shown to have increased the risk of continued infection and may improve functional outcome in patients with instability. Pyogenic vertebral osteomyelitis (PVO) is the most common spine infection in adults. It is acquired from a bacterial infection of the vertebral bodies that extends into or from adjacent intervertebral disc spaces. The annual incidence of PVO has been estimated at 0.059 episodes per 100,000 people over the past decade, with approximately 60% male predominance and a mean age of 66 years.1 The incidence of PVO is increasing, particularly in the immunocompomised and aging populations, according to two large national studies in Japan and Denmark. An increase in pathogen resistance to antibiotics may also be a contributing factor.2 There are various predisposing risk factors for spondylodiscitis, which include diabetes mellitus, rheumatic diseases, immunosuppressive diseases, previous invasive treatments (pharyngeal surgery, tonsillectomy, spine surgery, etc.), and sepsis or systemic infection (pneumonia, urinary tract infection, etc.).3 It is also increasingly recognized as a sequela of intravenous drug use, likely as a result of bacteremia. The recent increased prevalence of the disease results from a number of factors, including improved neuroimaging capabilities, better awareness of these diseases, an increased number of elderly individuals, and a greater number of surviving immunocompromised patients, including those with cancer, human immunodeficiency virus (HIV) infection. In addition, there is a growing rate of illicit drug abuse in the United States which may contribute as well as a greater incidence of methicillin-resistant Staphylococcus aureus (MRSA) infections. Despite having sophisticated imaging techniques and laboratory studies, the rate of misdiagnosis at presentation for vertebral osteomyelitis and epidural abscess is still high, with an average rate of approximately 50%.4 Vertebral osteomyelitis is notoriously difficult to initially diagnose because of its insidious course and indolent clinical course. Typical symptoms, such as back pain, tenderness, or weight loss, may not be prominent and can be present for months before a definitive diagnosis is made. The usual radiographic sequelae of these infections such as disc space narrowing and endplate destruction may not appear for weeks after the infection. Prompt imaging is still necessary for confirmation and localization of the infection.5 In a systematic review by Mylona et al of 1,008 patients with PVO and excluding patients with tuberculosis and brucellosis, they found the clinical picture of PVO often was nonspecific. Back pain was by far the most common presenting symptom. Notably, fever was often absent at presentation, thus distracting the clinician from the possibility of infection and delaying the diagnosis (mean time to diagnosis, 11–59 days).6 In these retrospective studies, the thoracic spine was affected 30% of the time, while the lumbar area was affected in 58% and cervical spine 11%. The highest rate of multifocal involvement was observed in patients with a history of intravenous drug abuse.6 The bacterial inoculation in PVO occurs through two main pathways: hematogenous seeding and direct inoculation. Batson et al described theories involving venous and arterial systems extensively, which describe the evolution of vertebral osteomyelitis. The Batson paravertebral plexus is a valveless venous system that allows retrograde flow of blood during times of increased intrathoracic or intra-abdominal pressure. This plexus is implicated in the spread of infection to central vertebral bodies, noncontiguous vertebral lesions, and seeding from distant sites (e.g., bacterial endocarditis).7 Bacteria enters areas of slower blood flow via arterial or venous conduits at the intervertebral disks or endplates, allowing extension into the vertebral column in 50% of cases.1 Bacteria enter the bloodstream and spread hematogenously to the subchondral area adjacent to the disc. The segmental arterial supply of the spine supplies two adjacent vertebras and the disc between them, and therefore, the infection involves two vertebral endplates as well as the disc. The infection spreads in several directions, further destroying the bone, and can eventually reach the spinal cord.5 Seeding of the spinal cord often occurs through arteries to the metaphysis of individual vertebrae, or less frequently by the Batson plexus or the deep pelvic venous system. This vascular spread of bacterial inoculation or pus can create an increase in intraosseous pressure that impedes blood flow to the vertebrae and intervertebral discs.8 A common source that must be excluded is bacterial endocarditis. Nearly 30% of all hematogenous spinal infections are associated with concomitant bacterial endocarditis. The bacteria S. aureus can cause not only a typical infection but also biomechanical instability and deformity by release of destructive enzymes such as hyaluronidase. This proteolytic enzyme can enhance the bacteria’s ability to invade connective tissue such as the disc’s annulus fibrosis, leading to the breakdown of its structural fibers, resulting in paraspinous disc extrusion.8 Bone destruction, ligamentous laxity, and nerve root/spinal cord compression leading to instability, deformity, and even severe neurologic deficits or death may result from the infiltration of these pathogenic organisms.1 Local inflammation from the response to infection may lead to thrombophlebitis or venous congestion extending to the subdural, and subarachnoid space is thought to be the mechanism of neurological injury in some cases. Surgical debridement is advocated by many as direct decompression and resection decreases possible mass effect and infectious burden, and also reduces the risk of vascular-related neurological complications.4 Routine invasive spinal procedures such as lumbar puncture and discography, and surgeries such as laminectomy, discectomy, and fusions can lead to direct inoculation causing bacterial colonization of the vertebral column. Direct inoculation is estimated to be causative in 15 to 40% of cases. The risk of post-procedural PVO are increased by prolonged operative time, instrumentation, posterior surgical approach, extensive soft-tissue dissection, and/or devitalization, creation of dead space, repeat surgery, surgery through previously irradiated tissue, excess blood loss, blood transfusions, and emergency surgery.9 Less commonly, bacterial inoculation occurs through local extension from adjacent areas of infection (3% of cases), including retropharyngeal abscesses or infected aortic grafts.1 Back pain is the most common symptom of osteomyelitis, accounting for the primary symptom in 67 to 100% patients from large studies (≥ 100 VO cases) and pooled data.9 Neck or back pain with insidious onset is seen in 85% of patients, and approximately 30% of patients present with concurrent neurological deficits. Other common nonspecific findings include fever (35–60%), weight loss, nausea/vomiting, anorexia, lethargy, or confusion. Tenderness of the spine on palpation is seen in 20% of patients with pyogenic vertebral osteomyelitis.7 Mylona et al systematically reviewed PVO clinical characteristics and reported the most common symptom is back pain (86%), with 34% of patients presenting with neurological deficits.6 Fevers are present in 85% of culture-positive PVO versus 32% of culture-negative cases. Concomitant infections (urinary tract infection, abscesses, skin infections, pneumonia, etc.) are relatively common, and are found in 47% of culture-positive and 4% of culture-negative PVO.1 There should be a high index of suspicion to prompt a diagnosis of PVO in the absence of fever and a non-specific insidious illness. Since the majority of PVO results from hematogenous seeding of an infection, the primary infection site, such as urinary tract or skin and soft tissue, often dominates the initial symptoms and signs.9 Standard laboratory workup may indicate an elevated white blood cell (WBC) count, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), blood culture, and urine cultures. An elevated ESR is the most reliable marker and is seen elevated in 90% of patients, while leukocytosis is seen in only 55% of VO cases, an elevated ESR is seen in 90% of patients. The CRP level, while also a nonspecific marker of inflammation, may be of more use in monitoring therapy because it has a shorter normalization time than ESR. In cases of VO, CRP level and ESR have been reported to have a sensitivity of 98 and 100%, respectively.7 Identification of pathogens, through either blood or tissue culture, is key to guiding antimicrobial therapy. Often a biopsy is required to identify the inciting organism and the antimicrobial sensitivities due to the low percentage of positive blood cultures in as little as 30% of cases.7 Mylona et al retrospectively reviewed fourteen cases of PVO, and in those cases blood cultures identified a pathogen in only 58% of the patients. When there were no positive blood cultures, when there was a failure to respond to antibiotics, or if a polymicrobial infection was suspected, a biopsy (CT-guided or open) was undertaken, providing a pathogen in 79% of the cases.6 Intraoperative tissue histopathology and cultures are the ideal reference tests for diagnosis of VO. However, they cannot be obtained in every patient with clinical and radiographic suspicion of VO. Therefore, the combination of multiple modalities along with clinical history and exam is reasonable for the diagnosis of VO. Image-guided percutaneous needle aspiration biopsy have a high positive predictive value for the diagnosis, but had moderate accuracy for ruling out this osteomyelitis. Image-guided percutaneous needle aspiration biopsy under computed tomography (CT) or fluoroscopic guidance has been recognized as a valuable method to obtain tissue diagnosis of vertebral disease, and the success rates of both methods are comparable. Its simplicity and cost-effectiveness led to the acceptance of this procedure as the standard tool for confirming a VO. Depending on the study, the diagnostic microbiological yields of percutaneous image-guided needle aspiration biopsy have been reported to vary from 36 to 91%.10 The wide range of success rates depend on the organism and multiple factors, such as prior use of antimicrobial therapy, biopsy techniques, and advances in imaging studies. A retrospective study by Pupaibool et al with seven retrospective studies involving 482 patients with clinical and/or radiological suspicion of VO who underwent image-guided spinal biopsy examined the diagnostic odds ratio (DOR), likelihood ratio of a positive test (LRP), likelihood ratio of a negative test (LRN), sensitivity, and specificity of image-guided biopsy. This review found that image-guided spinal biopsy had a DOR of 45.50 (95% confidence interval [CI], 13.66–151.56), an LRP of 16.76 (95% CI, 5.51–50.95), an LRN of 0.39 (95% CI, 0.24–0.64), a sensitivity of 52.2% (95% CI, 45.8–58.5), and a specificity of 99.9% (95% CI, 94.5–100).10 This study further confirmed the importance of image-guided biopsy for not only confirming the diagnosis but also guiding medical management with antibiotics. The differential diagnoses of PVO stem mostly from radiographic features which encompass pathologies such as erosive osteochondrosis, extruded disc material, vertebral fracture, metastatic disease, plasmacytoma, degenerative Modic’s changes, Charcot’s joint, and much more. These disease processes are not typically associated with elevated ESR, CRP, and WBC count, which are markers of inflammatory response. However, in a cancer or severely osteoporotic patient, the possibility of a pathological fracture should be included in the workup. With respect to an extruded disc fragment, some radiographic features such as T2 hyperintensity of an associated disc cysts and annulus tear associated with contrast enhancement make its radiographic presentation similar to PVO.11 This can lead to misdiagnosis as illustrated by Dunbar et al. Erosive osteochondrosis can involve the disc space, causing erosion of the vertebral body endplates and subsequent osteophytic and Modic’s changes in the vertebral bodies, which can have radiographic features similar to PVO. However, there often exists extension of the T2 signal beyond the affected endplate into the disc space in PVO. In addition to baseline age-associated degenerative changes in the spinal column, the destructive arthropathy caused by Charcot’s joint creates clinical and radiographic effects on the disc space and vertebral endplates with MRI features mimicking PVO.12 Plain radiographs have a lower sensitivity and specificity than cross-sectional imaging. However, X-rays provide important information regarding overall alignment and mechanical stability. Most suggestive findings on X-ray of osteomyelitis do not appear until later in the disease process, including osteolysis, endplate destruction, and eventual vertebral collapse.7 Evaluation of the entire spine should be considered in PVO, as 6% of patients demonstrate continuous lesions spanning multiple levels and 3% have noncontiguous, or skip lesions. While X-ray abnormalities for advanced cases of PVO can be found on nearly 90% of X-rays, early findings are nonspecific and difficult to identify. Early findings (2–3 weeks) may be endplate blurring, erosion of the endplate corners, disc height loss, and paraspinal soft-tissue swelling. Vertebral body destructive changes on X-ray are seen after greater than 30% bony destruction. Late PVO can demonstrate signs of bone formation, including peripheral sclerosis, osteophytosis, and osteolytic lesions.1 CT is the best imaging modality to evaluate instability, bony destruction, presence of gas within abscesses, and bony canal involvement. Visualization of calcifications within abscesses or identification of bone fragments can be accomplished with CT. Additionally, CT is the imaging modality used for image-guided biopsies in spinal infections. Findings of paraspinal calcifications and pedicle destruction are more common with granulomatous infections.7 CT may demonstrate epidural granulation tissue, possibly necessitating the need for closer characterization of the epidural space with MRI. Although CT scans can provide early information on bone integrity, they are limited when compared with MRI in identifying the extent of abscesses or spinal cord compression.1 MRI remains the imaging modality of choice when spinal infection is suspected. MRI helps to characterize the extent and location of the infection, including associated abscesses in the epidural and paraspinal compartments. It is also the most accurate modality to characterize the presence and extent of neural compression. MRI has a reported sensitivity of 96% and specificity of 94% in the diagnosis of spinal infection.7 Another recent series showed a sensitivity of 97.7% for detection of infection using MRI.13 MRI is useful in identifying some of the chronic changes that develop after development of PVO. MRI will most commonly demonstrate the contiguous involvement of two vertebrae with inflammatory change within the intervertebral disc ( Fig. 17.1). There is little evidence in the literature concerning the appearances of very early spinal infection on MRI and many times this must be inferred based on the clinical picture and ancillary imaging.13 Fig. 17.1 Imaging of a 56-year-old man with thoracic osteomyelitis. (a) Sagittal CT scan shows a destructive osteodiscitis at T7–T8. There is collapse of the disc space and a small focal kyphotic deformity. (b) Axial CT image through the T7 vertebral body demonstrates a patchy, lytic appearance to the body, which is infiltrated with osteomyelitis. (c) Sagittal MRI T1-weighted sequence with gadolinium contrast shows diffuse enhancement in the T7–T8 vertebral bodies as well as prespinal soft-tissue enhancement. There is a large, ventral epidural abscess behind the involved bodies, which is compressing the thoracic spinal cord. Affected discs and adjacent vertebrae may demonstrate high T2 and low T1 signal intensity early in the disease. STIR (short tau inversion recovery) sequence may demonstrate high T2 signal changes in the paraspinal soft tissues as well as inflammation and edema. Contrast is often used to display the diffuse enhancement of the subchondral bone and disc. The early MRI findings of high T2 disc signal with disc height loss and contrast uptake within the disc are highly sensitive (70–100%) for PVO diagnosis.1 However, Carragee and Iezza reviewed 103 MRIs of patients eventually diagnosed with PVO, and found a missed diagnosis rate of 9.1% with MRIs obtained within 2 weeks of patient presentation compared with a 3.4% missed diagnosis rate for MRIs after 2 weeks.14 An important diagnostic consideration in patients with pathological appearing vertebrae is ruling out neoplastic disease. This is usually done with laboratory tests that suggest infectious etiology, as previously discussed along with fever, and may be supported by MRI findings including disc space involvement with endplate erosion. It is common for neoplastic disease to spare the disc spaces, which sometimes can be a useful finding to distinguish these two entities. As VO progresses, further bony destruction is evident, producing a mass lesion that may cause cord compression. T1-weighted images show hypointensity in the disc and adjacent vertebral bodies, and T2-weighted images show hyperintensity in the same area. There is a loss of the margin between the involved disc and the adjacent vertebral bodies.5 MRI with weighted postcontrast imaging may demonstrate an abscess which would appear hypointense with characteristic peripheral rim enhancement. Dunbar et al presented several cases that confirm MRI images may be equivocal early in the course of infection. It is imperative where a history is consistent with VO, particularly if supported by positive blood cultures, that intravenous antibiotic therapy is continued and a repeat MRI performed.13
17.1 Introduction
17.1.1 Epidemiology
17.1.2 Clinical Features
17.1.3 Workup
17.1.4 Differential Diagnosis
17.1.5 Radiographic Appearance