Candida Spondylitis: Comparison of MRI Findings With Bacterial and Tuberculous Causes

Musculoskeletal Imaging Original Research Lee et al. MRI of Candida Spondylitis Musculoskeletal Imaging Original Research Sheen-Woo Lee 1 Sang Hoon Lee 2 Hye Won Chung 2 Min Jee Kim 3 Min Jeong Seo 4 Myung Jin Shin 2 Lee SW, Lee SH, Chung HW, Kim MJ, Seo MJ, Shin MJ Keywords: Candida, fungus, infectious spondylitis, MRI, opportunistic infection DOI:10.2214/AJR.12.10344 Received November 20, 2012; accepted after revision February 1, 2013. Presented at the 2008 annual meeting of the Radiological Society of North America. 1 Department of Radiology, Gachon University Gil Medical Center, Inchoen, Korea. 2 Department of Radiology, Asan Medical Center, University of Ulsan College of Medicine, 388-1 Pungnap-2 dong, Songpa-gu, Seoul, Korea. Address correspondence to S. H. Lee (shlee@amc.seoul.kr). 3 Department of Radiology, Seoul Medical Center, Seoul, Korea. 4 Department of Radiology, Seoul Veteran s Hospital, Seoul, Korea. AJR 2013; 201:872 877 0361 803X/13/2014 872 American Roentgen Ray Society Candida Spondylitis: Comparison of MRI Findings With Bacterial and Tuberculous Causes OBJECTIVE. Candida spondylitis is relatively uncommon and is usually encountered as an opportunistic infection. We analyzed the MRI characteristics of biopsy-proven cases of Candida spondylitis, and compared the findings with bacterial or tuberculous spondylitis. MATERIALS AND METHODS. The study included patients with infectious spondylitis who underwent MRI and biopsy from 1998 to 2011 (60 patients; mean age 56 ± 18 years). MR images were analyzed with respect to the number of involved vertebrae, contrast enhancement pattern, signal intensity of spinal inflammatory masses on T2-weighted imaging, paraspinal abscess size, intervertebral disk destruction, subligamentous spread, and skip lesions. The Fisher exact test and analysis of variance were used for statistical analysis. RESULTS. There were 10 cases of Candida spondylitis, and 29 and 21 cases of bacterial and tuberculous spondylitis, respectively. On MRI, disk destruction was seen in 50%, 93%, and 30% of Candida, bacterial, and tuberculous cases, respectively. Subligamentous spread of infection was noted in 22%, 10%, and 85%. Paraspinal inflammatory masses were seen in 100%, 100%, and 76%, and abscesses in 100%, 66%, and 90%, of Candida, bacterial, and tuberculous cases, respectively. Paraspinal inflammatory masses contained low T2 signal intensity portions in 80%, 21%, and 67%, and skip lesions were seen in 0%, 10%, and 14%, respectively. Small abscesses were noted in 100%, 76%, and 35% of Candida, bacteria, and tuberculosis infections, respectively. Candida involved 2.3 ± 0.4 vertebrae compared with 2.3 ± 0.9 and 3.0 ± 1.7 in bacterial and tuberculous, respectively. Differences in the three groups were statistically significant (p < 0.05) except for the number of involved vertebrae, and skip lesions. CONCLUSION. Candida spondylitis can be suspected when infectious lesions contain low-signal spinal inflammatory masses on T2-weighted imaging, small paraspinal abscesses, and in immunocompromised patients. T he risk of fungal infection is closely related with the immunocompromised state [1]. Although the CT and MRI findings of pyogenic or tuberculous spondylitis are well known, those of fungal spondylitis are not well characterized in the literature [2]. Fungal spondylitis is difficult to diagnose clinically or pathologically, and thus, treatment is commonly delayed [1]. Fungal spondylitis occurs primarily as an opportunistic infection [1], and Candida and Aspergillus species are the most common causative fungal organisms. Few cases of Candida spondylitis have been reported and little is known of its MRI characteristics. Furthermore, no study has systematically compared the imaging characteristics of Candida spondylitis and those of bacterial and tuberculous spondylitis. In this article, the MRI character- istics of biopsy-proven cases of Candida infection are analyzed and compared with those of the other more common causes of infectious spondylitis. Our objective was to identify the MRI characteristics of Candida spondylitis, and thus aid in its diagnosis. Materials and Methods With approval from our institutional review board, we retrospectively studied 352 patients who were admitted into our institute for infectious spondylitis from 1998 to 2011. Medical records were reviewed and clinical data and histopathologic findings were investigated. The inclusion criteria were as follows: prebiopsy MRI, including T1- and T2-weighted sagittal and axial imaging and contrast-enhanced studies; CT-guided percutaneous needle biopsy or open surgical biopsy of the infected spine area; and a positive pathogen yield by the biopsy. Patients whose medical record was unavailable; those without MRI; 872 AJR:201, October 2013

MRI of Candida Spondylitis or those with limited MR image quality, such as shading artifact by spine coil or who had a negative result on the direct biopsy, were excluded. The final study cohort consisted of 60 patients (23 women and 37 men) of mean age 56 ± 18 years. The MRI systems and pulse sequences used were varied because some of the imaging studies were performed at outside institutions, but all studies were performed using a 1.5-T magnet. Imaging sequences included spin-echo T1-weighted, T2-weighted without fat suppression, and gadolinium-enhanced T1-weighted with or without fat suppression. Imaging included axial and sagittal planes in all cases. Two musculoskeletal radiologists with 12 and 4 years of experience analyzed the MR images by consensus. The following MRI data were evaluated for each patient: the locations of involved bones, number of involved vertebra, presence of intervertebral disk destruction, subligamentous spread of inflammation, skip lesions, contrast enhancement pattern of the paraspinal soft-tissue lesion, signal intensity of the spinal inflammatory mass by T2- weighted imaging, and paraspinal abscess size. Intervertebral disk destruction was defined as breach of the subchondral cortex by extension of the infectious lesion through the endplate into the disk, with loss of the disk height on sagittal images. Subligamentous spread of inflammation was defined as an inflammatory mass or abscess under the anterior longitudinal ligament. Skip lesion was defined as spondylitis at different vertebral levels without intervening level involvement. The numbers of involved vertebrae were determined by counting involved vertebral bodies. Contrast enhancement patterns of the paraspinal soft-tissue lesions were assessed by reviewing the axial set of T1-weighted, T2-weighted, and contrast-enhanced images. The soft-tissue lesions outside the vertebral body showing low signal intensity on T1-weighted imaging and diffuse enhancement were categorized as paraspinal inflammatory masses, and those with a rim-enhancing component were categorized as abscesses. Further analyses of the paraspinal inflammatory mass and abscess were done in the following manner. Paraspinal inflammatory lesions were analyzed and categorized as diffusely high or heterogeneously low signal intensity compared with the adjacent muscle on axial T2-weighted imaging. The abscess size was categorized into large or small according to whether the individual cavity exceeded half of the diameter of the vertebral body. MR images were also reviewed twice by a radiologist unaware of causative organisms. Statistical differences between the three pathogenically different spondylitis types were analyzed using the Fisher exact test and analysis of variance. The correlation coefficient (R 2 ) was calculated to find out the possibility of association between the low signal intensity on T2-weighted imaging and the symptom duration or the patient s age. Analysis was performed using SPSS for Windows, version 12.0. Statistical significance was accepted for p values of < 0.05. TABLE 1: Underlying Diseases of Patients With Candida Spondylitis Patient No. Underlying Disease 1 Paraplegia due to bursting fracture of thoracic spine 2 Bedridden due to intestinal adhesion and fracture after motor vehicle accident 3 Chemotherapy due to lung cancer 4 After colon cancer operation 5 After colon cancer operation 6 After fusion and revision for spinal stenosis 7 Hepatocellular carcinoma, alcoholic liver cirrhosis 8 After diskectomy 9 Esophagus and lung cancer 10 Spine anesthesia for bladder tumor removal, 20-year history of diabetes mellitus Results There were 10 cases of Candida spondylitis, 29 cases of bacterial spondylitis, and 21 cases of tuberculous spondylitis. The cervical spine was not involved in any patients with Candida spondylitis or tuberculous spondylitis but was involved in two patients with bacterial spondylitis. The thoracic spine was involved in one Candida, three bacterial, and five tuberculous spondylitis patients. The thoracolumbar junction was involved in one bacterial and one tuberculous spondylitis patient, and the lumbar spine was involved in 10 Candida, 21 bacterial, and 16 tuberculous spondylitis patients. A review of clinical data showed that 90% of the Candida spondylitis patients were immunocompromised or had other long-standing debilitating disease due to cancer treatment or prolonged hospitalization associated with burst fracture or intestinal adhesion (Table 1). On the other hand, 10% of the pyogenic spondylitis patients and none of the tuberculous spondylitis patients had such conditions. Ten percent of the Candida spondylitis patients had undergone a recent invasive spinal procedure, such as, epidural injection or spine surgery, whereas 48% and 5% of pyogenic and tuberculous spondylitis patients, respectively, had done so. The respective durations of symptoms were 21 ± 44 months (median, 4 months), 2.0 ± 2.0 months (median, 1 month), and 14 ± 26 months (median, 5 months) in Candida and bacterial and tuberculous spondylitis patients, respectively. Mean patient age was 60 years (45 73 years) in the Candida spondylitis group, 60 years (15 82 years) in the bacterial spondylitis group, and 47 years (12 83 years) in the tuberculous spondylitis group. On MRI, intervertebral disk destruction was observed in 50%, 93%, and 30% of Candida, bacterial, and tuberculous cases (Fig. 1), and subligamentous spread of infection was noted in 22%, 10%, and 85% of Candida, bacterial, and tuberculous cases, respectively. A skip lesion was present in 10% of pyogenic and 14% of tuberculous spondylitis cases. Candida spondylitis involved 2.3 ± 0.4 vertebrae compared with 2.3 ± 0.9 in bacterial spondylitis and 3.0 ± 1.7 in tuberculosis. Diffusely enhancing paraspinal inflammatory masses were seen in 100%, 100%, and 76%, and rim-enhancing abscesses in 100%, 66%, and 90%, of Candida, bacterial, and tuberculous cases, respectively. Low-signal-intensity paraspinal mass on T2- weighted imaging was noted in 80%, 21%, and 67% of Candida, bacterial, and tuberculous cases, respectively (Fig. 2), whereas T2- high-signal-intensity paraspinal lesions were seen in 20%, 79%, and 24%, respectively (Fig. 2). Large confluent paraspinal abscesses were seen in 24% and 65% of bacterial and tuberculous cases, but none in Candida spondylitis (Figs. 2 and 3). Small abscesses were noted in 100%, 76%, and 35% of Candida, bacteria, and tuberculosis infections, respectively. The correlation coefficient (R 2 ) values between the T2 signal intensity and the symptom duration or the patient s age, calculated based on the hypothesis that the longer the symptom duration or the older the patient, the lower the T2 signal intensity, were 0.215 and 0.001, respectively. The low R 2 values suggest that the symptom duration or age does not cause significant effect on the T2 signal intensity. A summary of results is provided in Table 2. AJR:201, October 2013 873

Differences between the imaging findings in the three groups were statistically significant (p < 0.05) except for the number of involved vertebrae (p = 0.09) and presence of skip lesions (p = 0.54). Discussion The incidences of fungal infections have increased because of the greater use of immunosuppressants, prolonged use of broadspectrum antibiotics and indwelling catheters, and the higher prevalence of AIDS [1, 3]. Nevertheless, fungal spine infections still have much lower incidences than bacterial and tuberculous infections [4]. Knowledge of the disease itself is limited, and the diagnosis is based on exclusion rather than definite diagnostic criteria. In a clinical study, Frazier et al. [1] analyzed 11 cases in three tertiary centers over the course of 16 years, causative organisms included Candida, Aspergillus, and Coccidioides species. According to another researcher, the initial clinical manifestations of Candida infections are usually mild and nonspecific. This report concluded that early diagnosis of fungal spondylitis depends on a high index of suspicion and clinical judgment, especially in patients with a history of immunosuppression or of travel to endemic areas [5]. MRI can aid the diagnosis, but radiologic findings have not been evaluated in previous studies. Published imaging studies of fungal spondylitis involve anecdotal reports in the form of case studies, and the largest report issued on the subject involved three cases and no comparison with other causes of infectious spondylitis [6]. In this study, we analyzed the MRI findings of Candida spondylitis, the most common fungal spondylitis in the immunocompromised population, and we compared these with the findings of pyogenic and tuberculous spondylitis. Our results indicate that Candida spondylitis should be suspected if MRI shows spondylitis in two adjacent vertebral bodies, with small paraspinal abscess, or with phlegmon, which on T2-weighted imaging is of lower signal intensity in contrast to the high signal intensity expected of inflammatory lesions in cases with a pyogenic or tuberculosis infection. Less intervertebral disk involvement may also suggest Candida rather than pyogenic spondylitis. The well-known differences between pyogenic and tuberculous spondylitis depend on the virulence of the pathogen, aggressive proteolysis, and host-immune reactions. Variable host interactions with bacteria of low virulence Lee et al. TABLE 2: Spectrum of Imaging Findings of Candida Spondylitis, Bacterial Spondylitis, and Tuberculous Spondylitis Spondylitis MRI Findings Candida Bacterial Tuberculous Intervertebral disk destruction 50 93 30 Subligamentous spread 22 10 85 Skip lesion 0 10 14 Contrast enhancement pattern of paraspinal lesion Diffuse enhancing inflammatory mass 100 100 76 Rim-enhancing abscess 100 66 90 Low T2 signal intensity of inflammatory mass 80 21 67 No. of involving vertebrae (p = 0.09) 2.3 2.3 3.0 Large paraspinal abscess 0 24 65 Small paraspinal abscess 100 76 35 Note Findings with statistical significance (p < 0.05) are in bold. All data are presented as percentages except for number of involving vertebrae. or weaker proteolysis will result in atypical forms of spondylitis that can mimic tuberculosis or even a metastatic mass. For example, if the pathogen is low in virulence or the hostimmune reaction is low, the inflammatory tissue reaction will be low grade and long-standing, resulting in fibrosis in the inflammatory mass and a low tissue-water component, and the resulting signal intensity will be low on T2- weighted imaging [7 9]. The pathologic diagnoses of the cases in our study described chronic inflammation and fibrosis and support our hypothesis. Hence, chronic inflammation and fibrosis in Candida spondylitis and formation of paraspinal abscess cavities can resemble tuberculous spondylitis in some cases (Fig. 3). In addition, some species of fungus have been reported to synthesize paramagnetic substances, such as, melanin (a T2-shortening substance), which contributes to low signal intensity [10]. The list of pathologic reports of our Candida spondylitis patients included chronic osteomyelitis, dense fibrous tissue with calcifications, chronic granulomatous inflammation with necrosis, fibrinous exudate, marrow fibrosis with mild chronic inflammation, degenerated and necrotic synovial tissue, extensive degeneration of cartilage with granulation tissue, and chronic granulomatous inflammation with necrosis. In addition, underlying degenerative osteophytosis may contribute to the T2 low signal intensities of lesions in elderly patients. But the R 2 value in our study between the patients age and the T2 hypointensity, which can be seen in degenerative spur formation, did not indicate any relevance, suggesting that the effect of degenerative spine lesions may not be as significant as the hypothesis. Further study may be necessary in larger number of groups comparing CT and MRI side by side. The limitations of our study are as follows: Only patients with a biopsy-proven pathogen were included in this study. According to other researchers, infectious agents are found by biopsy in fewer than 70% of infectious spondylitis cases [11, 12]. In the current study, patients diagnosed with infectious spondylitis by imaging without a definite pathogen yield were treated with empirical antimicrobials, and exclusion of these patients might have led to selection bias and a higher proportion of atypical and severe cases. The cases in our study did not involve the cervical spine, but our literature search shows that cervical spine is not immune to Candida infection albeit unusual [1, 13 15]. The study was retrospective in nature, and some of the clinical information on the medical record, such as the history of antibiotic therapy before the visit to our tertiary medical center, was incomplete. Furthermore, although our cases were collected over 10 years, the sample size was small. In conclusion, Candida spondylitis should be considered when the infectious lesions involve contiguous vertebrae without intervertebral disk destruction with paraspinal inflammatory mass of unusually low signal intensity, and paraspinal abscess, when present, is small. A clinical history of chronic illness would be helpful during the differential diagnosis. These MRI findings may alert clinicians to the possibility of a rare cause of spondylitis and thus aid antimicrobial selection and patient management. 874 AJR:201, October 2013

MRI of Candida Spondylitis References 1. Frazier DD, Campbell DR, Garvey TA, Wiesel S, Bohlman HH, Eismont FJ. Fungal infections of the spine: report of eleven patients with long-term follow-up. J Bone Joint Surg Am 2001; 83-A:560 565 2. Johnson MD, Perfect JR. Fungal infections of the bones and joints. Curr Infect Dis Rep 2001; 3:450 460 3. Gamaletsou MN, Kontoyiannis DP, Sipsas NV, et al. Candida osteomyelitis: analysis of 207 pediatric and adult cases (1970 2011). Clin Infect Dis 2012; 55:1338 1351 4. Gouliouris T, Aliyu SH, Brown NM. Spondylodiscitis: update on diagnosis and management. J Antimicrob Chemother 2010; 65(suppl 3):iii11 iii24 5. Kim CW, Perry A, Currier B, Yaszemski M, Garfin SR. Fungal infections of the spine. Clin Orthop Relat Res 2006; 444:92 99 6. Williams RL, Fukui MB, Meltzer CC, Swarnkar A, Johnson DW, Welch W. Fungal spinal osteomyelitis in the immunocompromised patient: MR findings in three cases. AJNR 1999; 20:381 385 7. Hsu CY, Yu CW, Wu MZ, Chen BB, Huang KM, Shih TT. Unusual manifestations of vertebral osteomyelitis: intraosseous lesions mimicking metastases. AJNR 2008; 29:1104 1110 8. Tins BJ, Cassar-Pullicino VN, Lalam RK. Magnetic resonance imaging of spinal infection. Top Magn Reson Imaging 2007; 18:213 222 9. Ahmadi J, Bajaj A, Destian S, Segall HD, Zee CS. Spinal tuberculosis: atypical observations at MR imaging. Radiology 1993; 189:489 493 10. Doering TL, Nosanchuk JD, Roberts WK, Casadevall A. Melanin as a potential cryptococcal defence against microbicidal proteins. Med Mycol 1999; 37:175 181 Fig. 1 Candida spondylitis. A, MR images (T2-weighted, T1-weighted, and contrast-enhanced, respectively) show Candida spondylitis infection in lumbar spine in 68-year-old man, with little intervertebral disk destruction (arrow). B, MR images (T2-weighted, T1-weighted, and contrast-enhanced, respectively) show Candida spondylitis infection in 59-year-old man with disk destruction (arrowhead). Intervertebral disk was destroyed in 50% of Candida spondylitis patients compared with 93% of bacterial spondylitis patients and 28% of tuberculous spondylitis patients. 11. Vinicoff PG, Gutschik E, Hansen SE, Karle A, Rieneck K. CT-guided spinal biopsy in spondylodiscitis [in Danish]. Ugeskr Laeger 1998; 160:5931 5934 12. Rieneck K, Hansen SE, Karle A, Gutschik E. Microbiologically verified diagnosis of infectious spondylitis using CT-guided fine needle biopsy. APMIS 1996; 104:755 762 13. Lee DG, Park KB, Kang DH, Hwang SH, Jung JM, Han JW. A clinical analysis of surgical treatment for spontaneous spinal infection. J Korean Neurosurg Soc 2007; 42:317 325 14. Khazim RM, Debnath UK, Fares Y. Candida albicans osteomyelitis of the spine: progressive clinical and radiological features and surgical management in three cases. Eur Spine J 2006; 15:1404 1410 15. Müller EJ, Russe OJ, Muhr G. Osteomyelitis of the spine [in German]. Orthopade 2004; 33:305 315 A B AJR:201, October 2013 875

Lee et al. Fig. 2 Candida spondylitis. A C, MR images (T2-weighted, T1-weighted, and contrast-enhanced, respectively) of Candida spondylitis in 68-year-old man (A), bacterial spondylitis in 15-year-old boy (B), and tuberculous spondylitis in 41-year-old man (C). Axial images of Candida spondylitis show inflammatory mass in paraspinal soft tissue of mixed low signal intensity (thick white arrow, A) interspersed with punctate nonenhancing component (thin white arrow, A) compared with bacterial paraspinal inflammatory mass that is high signal intensity on T2-weighted image and diffusely enhanced (arrowhead, B) or thin rim-enhancing soft tissue abscess cavity of tuberculous spondylitis (curved arrow, C). A B C 876 AJR:201, October 2013

MRI of Candida Spondylitis Fig. 3 Paraspinal abscess and mass. A, MR images (T2-weighted, T1-weighted, and contrast-enhanced, respectively) of 68-year-old man with Candida spondylitis show multiple small paraspinal abscesses (thin arrow) and diffusely enhancing left paraspinal inflammatory mass that is low signal intensity on T2-weighted image (thick arrow). B, MR images (T2-weighted, T1-weighted, and contrast-enhanced, respectively) of 15-year-old boy with bacterial spondylitis show paraspinal inflammatory mass that is high signal intensity on T2-weighted image (double arrows). C and D, MR images (T2-weighted, T1-weighted, and contrast-enhanced, respectively) of 68-year-old man (C) and 27-year-old man (D) with tuberculous spondylitis show multiple small abscesses (black arrowhead) with T2-hypointense inflammatory mass (white arrowhead, C) and large confluent paraspinal abscess (curved arrow, D). A B C D AJR:201, October 2013 877