The Spine Journal
-
(2015)
-
Clinical study
Differential diagnosis between tuberculous spondylodiscitis and pyogenic spontaneous spondylodiscitis: a multicenter descriptive and comparative study Young K. Yoon, MDa, Yu M. Jo, MDb, Hyun H. Kwon, MDc, Hee J. Yoon, MDd, Eun J. Lee, MDe, So Y. Park, MDf, Seong Y. Park, MDg, Eun J. Choo, MDh, Seong Y. Ryu, MDi, Mi S. Lee, MDj, Kyung S. Yang, PhDk, Shin W. Kim, MDl,* a
Division of Infectious Diseases, Department of Internal Medicine, Korea University Anam Hospital, 3 Inchon-ro, Seongbuk-gu, Seoul 136-705, Republic of Korea b Division of Infectious Diseases, Department of Internal Medicine, Konyang University Hospital, 685 Gasuwon-dong seo-gu, Metropolitan city Daejon 302-718 Republic of Korea c Division of Infectious Diseases, Department of Internal Medicine, Daegu Catholic University Medical Center, 33 Duryungwon-ro Nam-gu, Daegu 705-718 Republic of Korea d Division of Infectious Diseases, Department of Internal Medicine, Eulji University Daejeon Hospital, 68 Hanglbisuk-ro, Nowon-gu, Seoul 139-872 Republic of Korea e Division of Infectious Diseases, Department of Internal Medicine, Soonchunhyang University Seoul Hospital, 59 Daesangwan-ro, Yongsan-gu, Seoul 140-887 Republic of Korea f Division of Infectious Diseases, Department of Internal Medicine, Kangdong Sacred Heart Hospital, 55, Beodeunaru-ro, Yeongdeungpo-gu, Seoul 150-037 Republic of Korea g Division of Infectious Diseases, Department of Internal Medicine, Dongguk University Ilsan Hospital, 27 Dongguk-ro Ilsandong-gu, Goyang-si, Gyeonggi-do 410-773 Republic of Korea h Division of Infectious Diseases, Department of Internal Medicine, Soonchunhyang University Bucheon Hospital, 170 jonaru-ro Wonmi-gu Bucheon city, Gyenggi-do 420-767 Republic of Korea i Division of Infectious Diseases, Department of Internal Medicine, Keimyung University Dongsan Hospital, 56 Dalseong-Ro, Jung-Gu, Daegu 700-712 Republic of Korea j Division of Infectious Diseases, Department of Internal Medicine, Kyung Hee University Medical Center, 23 Kyungheedae-ro Dongdaemun-gu Seoul 130-872, Republic of Korea k Department of Biostatistics, Korea University College of Medicine, 73 Inchon-ro, Seoul, Seoul 136-705, Republic of Korea l Division of Infectious Diseases, Department of Internal Medicine, Kyungpook National University Hospital, 130 Dongdoek-ro Jung-gu, Daegu 700-721, Republic of Korea Received 26 September 2014; revised 9 March 2015; accepted 2 April 2015
Abstract
BACKGROUND CONTEXT: Although tuberculous and pyogenic spondylodiscitis are common causes of spinal infections, their protean manifestation complicates differential diagnosis. PURPOSE: The clinical, laboratory, and radiologic characteristics of tuberculous and pyogenic spontaneous spondylodiscitis were compared in this study. STUDY DESIGN: This multicenter retrospective study was conducted in 11 teaching hospitals in the Republic of Korea from January 2011 to December 2013. PATIENT SAMPLE: Study subjects included adult patients ($18 years) diagnosed with tuberculous (n560) or pyogenic (n5117) spontaneous spondylodiscitis. OUTCOME MEASURES: Risk factors for tuberculous spondylodiscitis were determined, and their predictive performance was evaluated. METHODS: Multivariate logistic regression analysis was performed to determine predictors independently associated with tuberculous spondylodiscitis. Receiver-operating characteristic curve
FDA device/drug status: Not applicable. Author disclosures: YKY: Nothing to disclose. YMJ: Nothing to disclose. HHK: Nothing to disclose. HJY: Nothing to disclose. EJL: Nothing to disclose. SoYP: Nothing to disclose. SeYP: Nothing to disclose. EJC: Nothing to disclose. SYR: Nothing to disclose. MSL: Nothing to disclose. KSY: Nothing to disclose. SWK: Nothing to disclose. http://dx.doi.org/10.1016/j.spinee.2015.04.006 1529-9430/Ó 2015 Published by Elsevier Inc.
* Corresponding author. Division of Infectious Diseases, Department of Internal Medicine, Kyungpook National University Hospital, 130 Dongdoek-ro Jung-gu, Daegu 700-721, Republic of Korea. Tel.: (82) 53200-6525; fax: (82) 53-426-2046. E-mail address: [email protected] (S.W. Kim)
2
Y.K. Yoon et al. / The Spine Journal
-
(2015)
-
analysis using the presence or absence of risk factors was used to generate a risk index to identify patients with increased probability of tuberculous spondylodiscitis. RESULTS: Of 177 patients, multivariate logistic regression analysis showed that patients with tuberculous spondylodiscitis (n560) were more frequently women, with increased nonlumbar spinal involvement and associated non-spinal lesions, delayed diagnosis, higher serum albumin levels, reduced white blood cell counts, and lower C-reactive protein and procalcitonin levels. Among 117 patients with pyogenic spondylodiscitis, the most frequent causative microorganism was Staphylococcus aureus (64.1%). The mean diagnostic delay was significantly shorter, which may reflect higher clinical expression leading to earlier diagnosis. A combination of clinical data and biomarkers had better predictive value for differential diagnosis compared with biomarkers alone, with an area under the curve of 0.93, and sensitivity, specificity, and positive and negative predictive values of 95.0%, 79.5%, 70.4%, and 96.9%, respectively. CONCLUSIONS: This study provides guidance for clinicians to predict the causative organisms of spondylodiscitis in uncertain situations and before culture or pathologic examinations. Clinical data and single biomarkers combined can be useful for differential diagnoses between tuberculous and pyogenic spondylodiscitis. Ó 2015 Published by Elsevier Inc. Keywords:
Pyogenic; Tuberculosis; Spondylitis; Discitis; Leukocytes; Procalcitonin; C-reactive protein
Introduction Spondylodiscitis encompasses a wide range of vertebral osteomyelitis, spondylitis, and discitis; its increasing incidence is likely because of both an increasingly susceptible population and availability of advanced diagnostic tools [1–3]. The most severe spondylodiscitis complications include death and severe functional sequelae in 0% to 11% [4–10] and 5% to 35% [11–14] of patients, respectively. Delayed diagnosis and treatment is an independent risk factor for these adverse outcomes [7]. Because a variety of infectious pathogens can cause spondylodiscitis, it is difficult to determine its etiology without culture or pathologic examination. Therefore, specific microbiological diagnosis is preferred to confirm the infectious etiology and allow tailored antimicrobial therapy. However, the frequency of performing diagnostic biopsies varies (48%–100%) among spondylodiscitis studies [15]. Biopsy cultures from these series were positive in 43% to 90% of cases [5,7,16–19]. Therefore, diagnosis is difficult and often based on clinical, radiological, and laboratory features rather than microbiological tests. Knowledge of spondylodiscitis epidemiology is important to identify causative microorganisms globally, but this information has been very limited until recently. Most spinal infections are tuberculous or pyogenic. Previous studies in developed countries have reported tuberculous and pyogenic spondylodiscitis frequencies from 17% to 39% and 50% to 85%, respectively, among patients with spontaneous spinal infections [6,10,17,18,20–22]. Although differential diagnosis between tuberculous and pyogenic spondylodiscitis is essential for appropriate management, comparative studies of clinical, radiological, and laboratory differences to determine proper initial empirical antimicrobial therapies are limited [6,17,18,20]. The purpose of this multicenter comparative study was to describe the clinical characteristics of tuberculous and
pyogenic spondylodiscitis in the Republic of Korea, with particular focus on their differential diagnosis owing to the high tuberculosis burden. Materials and methods Study design and patients A multicenter retrospective comparative study was conducted in 11 teaching hospitals in the Republic of Korea from January 2011 to December 2013. Data were collected from the medical records of adult patients ($18 years) diagnosed with tuberculous or pyogenic spontaneous spondylodiscitis who were followed from treatment until death or end of treatment. This study included only patients diagnosed with spontaneous spondylodiscitis, except the patients with iatrogenic spondylodiscitis. Iatrogenic spondylodiscits was defined as spinal infection as a result of a penetrating injury or a surgical or radiological procedure. Only the first spondylodiscitis episodes were included in the analysis. The study protocol was approved before study initiation by the institutional review boards of each participating hospital. Definitions Pyogenic spondylodiscitis was diagnosed based on causative organisms isolated from spinal or paraspinal tissues and compatible symptoms, signs, or radiologic evidence of spondylodiscitis. Compatible spondylodiscitis symptoms or signs included spinal pain, limited motion, and tenderness over the involved spine, fever, or neurologic manifestations. Tuberculous spondylodiscitis was diagnosed when Mycobacterium tuberculosis was isolated from spinal or paraspinal tissues, smears were positive for acid-fast bacilli, tissue examination revealed caseating granuloma formation, or tissue samples were positive for
Y.K. Yoon et al. / The Spine Journal
M. tuberculosis on polymerase chain reaction (PCR), accompanied by compatible symptoms, signs, or radiologic evidence of spondylodiscitis. Candidate predictive factors were selected to guide the differential diagnosis between tuberculous and pyogenic spondylodiscitis based on the previous studies [6,17,20,23]. A relapse was defined as the reoccurrence of symptoms not attributable to other causes or new vertebral lesions on imaging together with elevated C-reactive protein (CRP) levels, increased erythrocyte sedimentation rates (ESRs), or both, after completing treatment. Diagnostic intervention to confirm the constancy of microbial diagnosis was not performed in the relapsed cases.
-
(2015)
-
3
CRP) and PCT levels that produce the best sensitivity and specificity based on the Youden index are not necessarily optimal in multivariate logistic regression analysis; therefore, a threshold analysis was performed to examine all possible combinations of cutoff values. Two-tailed p values less than .05 were considered statistically significant. All analyses were performed using IBM SPSS Statistics, version 20.0 (IBM Corporation, Armonk, NY, USA), R, version 2.15.2 (The R Foundation for Statistical Computing, Vienna, Austria), and SAS, version 9.2 (SAS Institute, Inc., Cary, NC, USA).
Results Variables
Patient characteristics
Data collected for this analysis included demographic characteristics, comorbidities, Charlson comorbidity index, [24] clinical manifestations, predisposing infection factors, time elapsed to diagnosis of spondylodiscitis, microbiological data, laboratory and radiologic data at onset, antimicrobial therapy, hospital stay, and inhospital mortality.
During the study period, data from 177 patients who met the inclusion criteria were collected from the participating hospitals. Patients were diagnosed with tuberculous spondylodiscitis (n560) or pyogenic spondylodiscitis (n5117). The same organism was identified in both tissue and blood cultures in 31 of 117 patients with pyogenic spondylodiscitis. Among 60 patients with tuberculous spondylodiscitis, M. tuberculosis was isolated from tissue samples from 50 patients; the remaining 10 patients were diagnosed based only on positive M. tuberculosis PCR results. Demographic and baseline characteristics of the 177 patients are listed in Table 1. Of these, 89 patients (50.3%) were women. The median age was 69 (IQR, 55–76) years. The proportion of women was significantly higher among patients with tuberculous spondylodiscitis compared with those with pyogenic spondylodiscitis. Univariate analyses revealed no significant age differences between the spondylodiscitis-positive groups (Table 1). The median Charlson comorbidity index was 2 (IQR, 0–2), and univariate analyses revealed that patients with pyogenic spondylodiscitis had significantly higher indices than those with tuberculous spondylodiscitis. Underlying cardiovascular diseases including hypertension, ischemic heart diseases, heart failure, and arrhythmia were significantly more common in patients with pyogenic spondylodiscitis (Table 1).
Statistical analyses Categorical and continuous variables were expressed as count (proportion) and median (interquartile range [IQR]), respectively. The Pearson chi-square test was used to compare groups of categorical variables. The Fisher exact test was used for expected case frequencies of five or less for any group. A two-sample Student t test was used to compare groups of normally distributed continuous variables. The Mann-Whitney U test was used to evaluate non-normal variable distributions. Receiver-operating characteristic (ROC) curves for each biomarker were derived and evaluated by comparing the area under the curves for assessing the accuracy of discriminable predictability between tuberculous and pyogenic spondylodiscitis, sensitivity, specificity, negative predictive value, and positive predictive value. Finally, the ROC curves of white blood cells (WBCs), CRP, and procalcitonin (PCT) were compared. Multivariate logistic regression analysis using the backward stepwise variable selection based on the Wald statistic was used to identify risk factors associated with tuberculous spondylodiscitis. Independent variables for multivariate logistic regression analysis were evaluated if they had at least 10% significance as predictors of tuberculous spondylodiscitis in the univariate analysis. The models were evaluated using the Hosmer-Lemeshow goodness-of-fit test. The predictive accuracy obtained by leave-one-out crossvalidation was used to evaluate the performance of a final logistic regression model. Receiver-operating curves were also constructed for the multivariate logistic regression model to evaluate discriminable predictability between tuberculous and pyogenic spondylodiscitis. The optimal cutoff values for individual biomarkers (i.e., WBC counts and
Clinical characteristics and predisposing factors The median diagnostic delay was 14 (IQR, 4–33) days and longer in patients with tuberculous spondylodiscitis (Table 2). Fever exceeding 38.0 C was present in 73 (62.4%) and 17 (28.3%) patients with pyogenic and with tuberculous spondylodiscitis, respectively (p!.001). There were no statistically significant differences in frequency of back pain and neurological symptoms between groups. Twenty-nine patients (16.4%) had accompanying meningitis, without significant difference in frequency between groups (Table 2).
4
Y.K. Yoon et al. / The Spine Journal
-
(2015)
-
Table 1 Comparison of demographic and baseline characteristics between patients with pyogenic or tuberculous spondylodiscitis Spondylodiscitis groups Variables*
All (n5177)
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
p Valuey
Female, n (%) Median age (y), median (IQR) Medical admission, n (%) Comorbidity, n (%) Cardiovascular disorders Central nervous system disorders Malignancy Renal disorders Hepatic disorders Respiratory disorders Connective tissue diseases Metabolic diseases Hematologic diseases Charlson comorbidity index, median (IQR)
89 (50.3) 69 (55–76) 93 (52.5)
71 (60.7) 68 (60–75) 71 (60.7)
18 (30.0) 69 (54–76) 22 (36.7)
!.001 .685 .002
61 17 18 9 15 12 5 45 7 0
47 12 15 7 31 7 4 35 5 1
14 5 3 2 2 5 1 10 2 0
(34.5) (9.6) (10.2) (5.1) (8.5) (6.8) (2.8) (25.4) (4.0) (0–2)
(40.2) (10.3) (12.8) (6.0) (11.1) (6.0) (3.4) (29.9) (4.3) (0–2)
(23.3) (8.3) (5.0) (3.3) (3.3) (8.3) (1.7) (16.7) (3.3) (0–1)
.026 .681 .103 .720 .079 .544 .663 .055 1.000 .005
IQR, interquartile range. * Values represent the number of subjects (%), mean (standard deviation), or median (IQR). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
Table 3 shows predisposing factors and non–spinal-associated lesions in patients with spontaneous spondylodiscitis. Non–spinal-associated lesions were more frequently associated with patients with tuberculous spondylodiscitis than those with pyogenic spondylodiscitis. Radiologic findings and laboratory data The most frequently involved region in both groups was the lumbar spine. However, nonlumbar involvement was significantly more frequent in patients with tuberculous spondylodiscitis (24 [40.0%] vs. 27 [23.1%], p5.019). Forty-three (24.3%) patients had involvement of three or more levels. Epidural or paraspinal abscess was present in 72 (61.5%) and 38 (63.3%) patients with pyogenic and tuberculous spondylodiscitis, respectively (p5.816). Table 2 summarizes laboratory findings in both groups. Compared with those with tuberculous spondylodiscitis, patients with pyogenic spondylodiscitis were more frequently associated with the following: hematocrit less than 30%, WBC count greater than or equal to 15,000/mm3, proportion of neutrophils greater than 75%, serum albumin level less than 3.0 g/dL, and CRP level greater than 7.25 mg/L (Table 2). Fig. 1 represents the differential diagnostic accuracy of the biomarkers for patients with spontaneous spondylodiscitis, estimated with the area under the curve (AUC) according to the cutoff values of the single biomarkers. Microbiological characteristics All 177 patients in this study underwent tissue examination. Among 117 patients with pyogenic spondylodiscitis, 2 had polymicrobial samples, with methicillin-resistant Staphylococcus aureus (MRSA), Streptococcus agalactiae, methicillin-susceptible S. aureus (MSSA), and Pseudomonas
aeruginosa. The other 115 pyogenic spondylodiscitis cases were monomicrobial, with the following isolated organisms: MSSA (n555), MRSA (n518), coagulase-negative Staphylococcus (n510), Streptococcus spp. (n511), Enterococcus faecalis (n51), Escherichia coli (n59), Klebsiella pneumoniae (n56), P. aeruginosa (n51), Serratia marcescens (n51), Citrobacter freundii (n51), Proteus mirabilis (n51), and Salmonella Group C (n51). Among 60 patients with tuberculous spondylodiscitis, M. tuberculosis was isolated from tissues in 50 patients, and 43 were for positive for M. tuberculosis on PCR. Treatment outcomes Table 4 shows treatment and clinical outcomes. The overall all-cause inhospital and spondylodiscitis-related mortalities were 5.6% (10/177) and 2.8% (5/177), respectively. There were no significant differences in all-cause inhospital and spondylodiscitis-related mortality between groups. The frequency of surgical intervention because of spondylodiscitis was higher among patients with tuberculous spondylodiscitis compared with those with pyogenic spondylodiscitis. After diagnosis of spontaneous spondylodiscitis, the median hospital stay was significantly longer in patients with pyogenic spondylodiscitis. Factors for differential diagnosis between groups Multiple logistic regression modeling revealed that the statistically significant factors associated with tuberculous spondylodiscitis included sex (female, odds ratio, 8.84; 95% confidence interval, 2.78–28.07), involvement of nonlumbar spine (3.48, 1.16–10.47), elapsed to time spondylodiscitis diagnosis more than 7 days (4.79, 1.58–14.51), WBC count less than or equal to 9,600/mm3 (29.82, 6.34– 140.17), a combination of serum albumin levels greater than
Y.K. Yoon et al. / The Spine Journal
-
(2015)
-
5
Table 2 Comparison of clinical characteristics between patients with pyogenic or tuberculous spondylodiscitis Spondylodiscitis groups All (n5177)
Variables* Clinical features, n (%) Time elapsed to admission (d), median (IQR) Time elapsed to diagnosis of spondylodiscitis #7 d Fever Back pain Neurological symptoms Meningitis, n (%) Involved vertebral regions, n (%) Cervical Thoracic Lumbar Sacral $3 spinal regions Skipping lesions Paraspinal abscess Epidural abscess Psoas muscle abscess Laboratory results at time of initial diagnosis, n (%) Anemia (hematocrit !30) Leukocyte count $15,000/mm3 Neutrophilia $75% Hypoalbuminemia !3.0 g/dL WBC count (per mm3), median (IQR) CRP level (mg/L), median (IQR) CRP level O7.25 mg/L ESR (mm/h), median (IQR) ALP level (IU/L), median (IQR) PCT level (ng/dL), median (IQR) Interferon-gamma release assays
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
14 118 90 149 77 29
(4–30) (66.7) (50.8) (84.2) (43.5) (16.4)
10 67 73 102 48 10
(4–29) (57.3) (62.4) (87.2) (41.0) (16.7)
30 51 17 47 29 19
(12–73) (85.0) (28.3) (78.3) (48.3) (16.2)
17 48 126 19 43 6 62 87 43
(9.6) (27.1) (71.2) (10.7) (24.3) (3.4) (35.0) (49.2) (24.3)
10 24 90 15 25 4 38 58 29
(8.5) (20.5) (76.9) (12.8) (21.4) (3.4) (32.5) (49.6) (24.8)
7 24 36 4 18 2 24 29 14
(11.7) (40.0) (60.0) (6.7) (30.0) (3.3) (40.0) (48.3) (23.3)
55 30 113 48 9,300 12.5 107 63 111 0.51 22/42
(31.1) (16.9) (63.8) (27.1) (6,970–13,450) (3.3–34.7) (60.5) (40–83) (60–216) (0.082–1.41) (52.4)
43 30 89 44 11,900 19.9 88 63 129.0 0.59 5/18
(36.8) (25.6) (76.1) (37.6) (7,650–15,560) (6.9–80.7) (75.2) (39–86) (68.8–217.0) (0.11–1.67) (27.8)
12 (20.0) 0 24 (40.0) 4 (6.7) 7,450 (5,655–8,882) 3.2 (1.8–12.3) 19 (31.7) 62 (40–80) 94 (31–205) 0.11 (0.06–0.70) 17/24(70.8)
p Valuey !.001 !.001 !.001 .127 .353 .942 .505 .006 .019 .211 .266 1.000 .321 .876 .831 .023 !.001 !.001 !.001 !.001 !.001 !.001 .519 .098 .066 .006
ALP, alkaline phosphatase; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; PCT, procalcitonin; WBC, white blood cell. * Values represent the number of subjects (%), mean (standard deviation), or median (interquartile range). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
or equal to 3.0 mg/dL and PCT levels less than or equal to 1.39 ng/dL (8.97, 1.65–48.69), and a combination of non– spinal-associated lesions and CRP levels #7.25 mg/L (8.59, 1.62–45.64) (Table 5). The p value for the HosmerLemeshow goodness-of-fit test was 0.954, greater than the 0.05 significance threshold; therefore, there is no significant evidence for lack of fit in any of the final models.
Leave-one-out cross-validation was performed to assess the predictive accuracy of each final model. The AUCs for the clinical failure model were about 0.90 for both raw data and leave-one-out cross-validation. For this model, the sensitivity, specificity, and positive predictive value and negative predictive value obtained with an optimal cutoff point were greater than 0.70 (Fig. 2).
Table 3 Comparison of predisposing factors between patients with pyogenic or tuberculous spondylodiscitis Spondylodiscitis groups Variables*
All (n5177)
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
p Valuey
Invasive spinal procedures Previous admission Previous operations Recent bacteremia Exposure to immunosuppressant History of tuberculosis Non–spinal-associated lesions
7 61 16 30 7 29 56
4 42 11 22 4 23 31
3 19 5 8 3 6 25
.690 .575 .814 .359 .690 .100 .040
(4.0) (34.5) (9.0) (16.9) (4.0) (16.4) (31.6)
(3.4) (35.9) (9.4) (18.8) (3.4) (19.7) (26.5)
(5.0) (31.7) (8.3) (13.3) (5.0) (10.0) (41.7)
* Values represent the number of subjects (%), mean (standard deviation), or median (interquartile range). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
6
Y.K. Yoon et al. / The Spine Journal
Fig. 1. Receiver-operating characteristic curve of white blood cell counts, C-reactive protein levels, and procalcitonin levels to differentiate patients with tuberculous or pyogenic spondylodiscitis.
Discussion This multicenter comparative study investigated the clinical, laboratory, and radiological characteristics of spontaneous spondylodiscitis secondary to tuberculosis and other bacterial causes. In this study, predictors useful for differential diagnosis between the spondylodiscitis groups before culture or pathologic examinations were identified, and ROC curve analysis using serum WBC counts and CRP and PCT levels produced a risk index to identify patients with an increased probability of tuberculous spondylodiscitis. Our findings suggest that a combination of information derived from clinical examination and laboratory parameters, such as WBC counts and CRP or PCT levels, may be useful for differential diagnosis between tuberculous and pyogenic spondylodiscitis. Tuberculosis is a global concern in developing and developed countries. Despite rapid economic development, its prevalence in the Republic of Korea is the highest among Organization for Economic Cooperation and Development nations, with an incidence of 108 cases per 100,000 people [25]. Patients with spinal tuberculosis are not
-
(2015)
-
common but comprise up to 50% of all patients afflicted with musculoskeletal tuberculosis. In tuberculosisendemic countries such as the Republic of Korea, a clinical suspicion of tuberculous spondylodiscitis is required for early selection of appropriate antimicrobial therapies. Data to guide the differential diagnosis between tuberculous and pyogenic spondylodiscitis have been identified in a few single-center studies with small sample sizes. Known risk factors for tuberculous spondylodiscitis included chronic renal failure, absence of fever, constitutional symptoms, neurological deficit, psoas abscess, history of tuberculosis, active tuberculosis of other organs, increased diagnostic delay, presence of spinal deformities, involvement of the thoracic spine or greater than or equal to three spinal levels, higher ESR, and involvement of posterior elements [6,17,20]. Risk factors for pyogenic spondylodiscitis included previous invasive spinal procedures, preceding bacteremia, diabetes mellitus, underlying chronic debilitating diseases, chronic renal failure, liver cirrhosis, intravenous drug abuse, immunosuppression, fever (temperature O38 C), WBC count greater than 1,000/mm3, neutrophil fractions greater than 75%, CRP level greater than 5 mg/ dL, ESR greater than 40 mm/h, and alkaline phosphatase level greater than 120 IU/L [6,20]. Several predictors identified in our study were comparable with the previous reports [17,20]. The mean diagnostic delay was significantly shorter among patients with pyogenic spondylodiscitis, which may reflect higher clinical expression that permits an earlier diagnosis. Although tuberculosis affects men more often than women, women were more frequently diagnosed with tuberculous spondylodiscitis, another finding similar to the other studies [20,25]. Previous studies reported lumbar and thoracic spines to be the most common regions of spondylodiscitis among patients with pyogenic and tuberculous spondylodiscitis, respectively [23]. However, we found the lumbar region to be most commonly affected region in both groups. Involvement of the thoracic or cervical spine was more frequently associated with tuberculous spondylodiscitis (Table 2). In this study, the 5.6% mortality rate among patients with pyogenic spondylodiscitis was similar to previously
Table 4 Comparison of treatment and clinical outcomes among patients with pyogenic and tuberculous spondylodiscitis Spondylodiscitis groups Variables* Antimicrobial treatment Duration of appropriate treatment (d), median (IQR) Operation because of spondylitis Outcomes, n (%) Treatment relapse Complications Hospital stay after onset of spondylitis (d), median (IQR) Inhospital mortality, n (%) Mortality attributable to spondylitis, n (%)
All (n5177)
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
p Valuey
76 (30–312) 102 (57.6)
56 (22–89) 58 (49.6)
360 (257–368) 44 (73.3)
!.001 .002
9 22 41 10 5
(5.1) (12.4) (26–65) (5.6) (2.8)
7 12 47 8 5
(6.0) (10.3) (34–67) (6.8) (4.3)
IQR, interquartile range. * Values represent the number of subjects (%), mean (standard deviation), or median (IQR). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
2 (3.3) 10 (16.7) 30 (17–55) 2 (3.3) 0
.720 .221 .001 .498 .168
Y.K. Yoon et al. / The Spine Journal Table 5 Multivariate logistic regression analyses of risk factors associated with tuberculous spondylodiscitis in 177 patients with pyogenic or tuberculous spondylodiscitis Independent variables* Female sex Involvement of nonlumbar spine Time elapsed to diagnosis of spondylodiscitis O7 d WBC count #9,600/mm3 Serum albumin level $3.0 mg/dL and PCT level #1.39 ng/dL Associated non-spinal lesions and CRP level #7.25 mg/L
OR (95% CI for OR)
p Value
8.84 (2.78–28.07) 3.48 (1.16–10.47) 4.79 (1.58–14.51)
!.001 .027 .006
29.82 (6.34–140.17) 8.97 (1.65–48.69)
!.001 .011
8.59 (1.62–45.64)
.012
CI, confidence interval; CRP, C-reactive protein; OR, odds ratio; PCT, procalcitonin; WBC, white blood cell. * Female sex, diabetes mellitus, Charlson comorbidity index, fever ($38 C), involvement of thoracic spine, involvement of nonlumbar spine, non–spinal-associated lesions, time elapsed to diagnosis of spondylodiscitis O7 days, hematocrit !30%, serum albumin $3.0 mg/dL, WBC count #9,600/mm3, CRP level #7.25 mg/L, and PCT level #1.39 ng/dL were evaluated as independent variables for multivariable logistic regression analysis if they were predictors of tuberculous spondylodiscitis at the 10% significance level.
reported rates of 4% to 10% [18,26,27]. No patients with tuberculous spondylodiscitis died, comparable with mortality rates of 0% and 7.3% reported by the previous studies and attributed to other causes [6,20]. Spondylodiscitis prognosis might depend on early diagnosis and treatment or identification of the causal agent [18,26,27]. In our study, the five mortality cases attributable to spondylodiscitis were pyogenic spondylodiscitis, caused by MRSA (n53), MSSA (n51), and Streptococcus viridans (n51), with an elapsed time to diagnosis of 14 days or more, except for a single patient. Similar to the previous reports, surgical
-
(2015)
-
7
treatment was more frequently associated with tuberculous than pyogenic spondylodiscitis [17]. Our study also found that WBC counts and CRP and PCT levels might be valuable laboratory findings to diagnose spondylodiscitis. Univariate analyses indicated that higher levels of these biomarkers were suggestive of pyogenic rather than tuberculous spondylodiscitis. The AUC for these three biomarkers was approximately 0.7 (Fig. 1). Multivariate logistic regression modeling showed that clinical information along with laboratory findings resulted in better differential diagnosis between spondylodiscitis groups compared with the discriminative ability of each individual biomarker (Fig. 2). This study has some limitations. First, this study was performed in a tuberculosis-endemic region; therefore, these results may not be generalizable to other populations or areas with low tuberculosis rates. Second, all subjects in this study were microbiologically confirmed patients. However, the causative microorganisms in clinical practice are often not identified microbiologically in patients with spondylodisictis [1,28]. The predictability of these risk factors to discriminate between patients with tuberculous and pyogenic spondylodiscitis was not evaluated in those with a negative culture. Further research is required to determine whether these predictive factors may be more helpful for deciding initial empirical antibiotics before the identification of the causative microorganism or for planning the antibiotic regimen in the patients who were not microbiologically confirmed. Third, the ranges in the quantitative indices of the laboratory biomarkers used in this study had a wide overlap between pyogenic and tuberculous spondylodiscitis. Therefore, it was difficult to determine an accurate cutoff value for discriminating each biomarker. However, the cutoff values in the multivariate logistic regression analyses were determined at the highest AUC value in conjunction with other predictive factors (Table 5). In conclusion, this multicenter comparative study suggests that a combination of clinical data and laboratory findings can be useful to discriminate between patients with tuberculous and pyogenic spondylodiscitis. Particularly, these predictive factors may be useful for selecting the initial empirical antibiotics before culture results are available. Further prospective randomized trials with larger study populations are necessary to develop better predictive models to guide selection of appropriate initial empirical antimicrobial therapies in patients with spondylodiscitis. Acknowledgments
Fig. 2. Receiver-operating characteristic curve for tuberculous spondylodiscitis obtained using the predictive probability of multivariate logistic regression model and validation results.
This work was supported in part by the Korean Society for Chemotherapy and by Roche, in the Republic of Korea. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Institutional review board approval/research ethics committee: The study protocol was approved before study
8
Y.K. Yoon et al. / The Spine Journal
initiation by the institutional review boards of each participating hospital. References [1] Lora-Tamayo J, Euba G, Narvaez JA, Murillo O, Verdaguer R, Sobrino B, et al. Changing trends in the epidemiology of pyogenic vertebral osteomyelitis: the impact of cases with no microbiologic diagnosis. Semin Arthritis Rheum 2011;41:247–55. [2] Cottle L, Riordan T. Infectious spondylodiscitis. J Infect 2008;56: 401–12. [3] Lew DP, Waldvogel FA. Osteomyelitis. N Engl J Med 1997;336: 999–1007. [4] Legrand E, Flipo RM, Guggenbuhl P, Masson C, Maillefert JF, Soubrier M, et al. Management of nontuberculous infectious discitis. Treatments used in 110 patients admitted to 12 teaching hospitals in France. Joint Bone Spine 2001;68:504–9. [5] Beronius M, Bergman B, Andersson R. Vertebral osteomyelitis in G€ oteborg, Sweden: a retrospective study of patients during 199095. Scand J Infect Dis 2001;33:527–32. [6] Colmenero JD, Jimenez-Mejıas ME, Sanchez-Lora FJ, Reguera JM, Palomino-Nicas J, Martos F, et al. Pyogenic, tuberculous, and brucellar vertebral osteomyelitis: a descriptive and comparative study of 219 cases. Ann Rheum Dis 1997;56:709–15. [7] McHenry MC, Easley KA, Locker GA. Vertebral osteomyelitis: longterm outcome for 253 patients from 7 Cleveland-area hospitals. Clin Infect Dis 2002;34:1342–50. [8] Friedman JA, Maher CO, Quast LM, McClelland RL, Ebersold MJ. Spontaneous disc space infections in adults. Surg Neurol 2002;57:81–6. [9] Wirtz DC, Genius I, Wildberger JE, Adam G, Zilkens KW, Niethard FU. Diagnostic and therapeutic management of lumbar and thoracic spondylodiscitis—an evaluation of 59 cases. Arch Orthop Trauma Surg 2000;120:245–51. [10] Nolla JM, Ariza J, Gomez-Vaquero C, Fiter J, Bermejo J, Valverde J, et al. Spontaneous pyogenic vertebral osteomyelitis in nondrug users. Semin Arthritis Rheum 2002;31:271–8. [11] Ross PM, Fleming JL. Vertebral body osteomyelitis: spectrum and natural history. A retrospective analysis of 37 cases. Clin Orthop Relat Res 1976;118:190–8. [12] Osenbach RK, Hitchon PW, Menezes AH. Diagnosis and management of pyogenic vertebral osteomyelitis in adults. Surg Neurol 1990;33:266–75. [13] Griffiths HE, Jones DM. Pyogenic infection of the spine. A review of twenty-eight cases. J Bone Joint Surg Br 1971;53:383–91.
-
(2015)
-
[14] Cahill DW, Love LC, Rechtine GR. Pyogenic osteomyelitis of the spine in the elderly. J Neurosurg 1991;74:878–86. [15] Mylona E, Samarkos M, Kakalou E, Fanourgiakis P, Skoutelis A. Pyogenic vertebral osteomyelitis: a systematic review of clinical characteristics. Semin Arthritis Rheum 2009;39:10–7. [16] Hadjipavlou AG, Mader JT, Necessary JT, Muffoletto AJ. Hematogenous pyogenic spinal infections and their surgical management. Spine 2000;25:1668–79. [17] Turunc T, Demiroglu YZ, Uncu H, Colakoglu S, Arslan H. A comparative analysis of tuberculous, brucellar and pyogenic spontaneous spondylodiscitis patients. J Infect 2007;55:158–63. [18] Perronne C, Saba J, Behloul Z, Salmon-Ceron D, Leport C, Vilde JL. Pyogenic and tuberculous spondylodiscitis (vertebral osteomyelitis) in 80 adult patients. Clin Infect Dis 1994;19:746–50. [19] Karadimas EJ, Bunger C, Lindblad BE, Hansen ES, Høy K, Helmig P. Spondylodiscitis. A retrospective study of 163 patients. Acta Orthop 2008;79:650–9. [20] Kim CJ, Song KH, Jeon JH, Park WB, Park SW, Kim HB, et al. A comparative study of pyogenic and tuberculous spondylodiscitis. Spine 2010;35:E1096–100. [21] Luzzati R, Giacomazzi D, Danzi MC, Tacconi L, Concia E, Vento S. Diagnosis, management and outcome of clinically-suspected spinal infection. J Infect 2009;58:259–65. [22] Belzunegui J, Del Val N, Intxausti JJ, De Dios JR, Queiro R, Gonzalez C, et al. Vertebral osteomyelitis in northern Spain. Report of 62 cases. Clin Exp Rheumatol 1999;17:447–52. [23] Lee KY. Comparison of pyogenic spondylitis and tuberculous spondylitis. Asian Spine J 2014;8:216–23. [24] Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol 1994;47:1245–51. [25] World Health Organization. Global tuberculosis control. Geneva, Switzerland: World Health Organization, 2013. [26] Buchelt M, Lack W, Kutschera HP, Katterschafka T, Kiss H, Schneider B, et al. Comparison of tuberculous and pyogenic spondylitis. An analysis of 122 cases. Clin Orthop Relat Res 1993;296: 192–9. [27] Kim J, Kim YS, Peck KR, Kim ES, Cho SY, Ha YE, et al. Outcome of culture-negative pyogenic vertebral osteomyelitis: comparison with microbiologically confirmed pyogenic vertebral osteomyelitis. Semin Arthritis Rheum 2014;44:246–52. [28] Bhagat S, Mathieson C, Jandhyala R, Johnston R. Spondylodiscitis (disc space infection) associated with negative microbiological tests: comparison of outcome of suspected disc space infections to documented non-tuberculous pyogenic discitis. Br J Neurosurg 2007;21: 473–7.
-
(2015)
-
Clinical study
Differential diagnosis between tuberculous spondylodiscitis and pyogenic spontaneous spondylodiscitis: a multicenter descriptive and comparative study Young K. Yoon, MDa, Yu M. Jo, MDb, Hyun H. Kwon, MDc, Hee J. Yoon, MDd, Eun J. Lee, MDe, So Y. Park, MDf, Seong Y. Park, MDg, Eun J. Choo, MDh, Seong Y. Ryu, MDi, Mi S. Lee, MDj, Kyung S. Yang, PhDk, Shin W. Kim, MDl,* a
Division of Infectious Diseases, Department of Internal Medicine, Korea University Anam Hospital, 3 Inchon-ro, Seongbuk-gu, Seoul 136-705, Republic of Korea b Division of Infectious Diseases, Department of Internal Medicine, Konyang University Hospital, 685 Gasuwon-dong seo-gu, Metropolitan city Daejon 302-718 Republic of Korea c Division of Infectious Diseases, Department of Internal Medicine, Daegu Catholic University Medical Center, 33 Duryungwon-ro Nam-gu, Daegu 705-718 Republic of Korea d Division of Infectious Diseases, Department of Internal Medicine, Eulji University Daejeon Hospital, 68 Hanglbisuk-ro, Nowon-gu, Seoul 139-872 Republic of Korea e Division of Infectious Diseases, Department of Internal Medicine, Soonchunhyang University Seoul Hospital, 59 Daesangwan-ro, Yongsan-gu, Seoul 140-887 Republic of Korea f Division of Infectious Diseases, Department of Internal Medicine, Kangdong Sacred Heart Hospital, 55, Beodeunaru-ro, Yeongdeungpo-gu, Seoul 150-037 Republic of Korea g Division of Infectious Diseases, Department of Internal Medicine, Dongguk University Ilsan Hospital, 27 Dongguk-ro Ilsandong-gu, Goyang-si, Gyeonggi-do 410-773 Republic of Korea h Division of Infectious Diseases, Department of Internal Medicine, Soonchunhyang University Bucheon Hospital, 170 jonaru-ro Wonmi-gu Bucheon city, Gyenggi-do 420-767 Republic of Korea i Division of Infectious Diseases, Department of Internal Medicine, Keimyung University Dongsan Hospital, 56 Dalseong-Ro, Jung-Gu, Daegu 700-712 Republic of Korea j Division of Infectious Diseases, Department of Internal Medicine, Kyung Hee University Medical Center, 23 Kyungheedae-ro Dongdaemun-gu Seoul 130-872, Republic of Korea k Department of Biostatistics, Korea University College of Medicine, 73 Inchon-ro, Seoul, Seoul 136-705, Republic of Korea l Division of Infectious Diseases, Department of Internal Medicine, Kyungpook National University Hospital, 130 Dongdoek-ro Jung-gu, Daegu 700-721, Republic of Korea Received 26 September 2014; revised 9 March 2015; accepted 2 April 2015
Abstract
BACKGROUND CONTEXT: Although tuberculous and pyogenic spondylodiscitis are common causes of spinal infections, their protean manifestation complicates differential diagnosis. PURPOSE: The clinical, laboratory, and radiologic characteristics of tuberculous and pyogenic spontaneous spondylodiscitis were compared in this study. STUDY DESIGN: This multicenter retrospective study was conducted in 11 teaching hospitals in the Republic of Korea from January 2011 to December 2013. PATIENT SAMPLE: Study subjects included adult patients ($18 years) diagnosed with tuberculous (n560) or pyogenic (n5117) spontaneous spondylodiscitis. OUTCOME MEASURES: Risk factors for tuberculous spondylodiscitis were determined, and their predictive performance was evaluated. METHODS: Multivariate logistic regression analysis was performed to determine predictors independently associated with tuberculous spondylodiscitis. Receiver-operating characteristic curve
FDA device/drug status: Not applicable. Author disclosures: YKY: Nothing to disclose. YMJ: Nothing to disclose. HHK: Nothing to disclose. HJY: Nothing to disclose. EJL: Nothing to disclose. SoYP: Nothing to disclose. SeYP: Nothing to disclose. EJC: Nothing to disclose. SYR: Nothing to disclose. MSL: Nothing to disclose. KSY: Nothing to disclose. SWK: Nothing to disclose. http://dx.doi.org/10.1016/j.spinee.2015.04.006 1529-9430/Ó 2015 Published by Elsevier Inc.
* Corresponding author. Division of Infectious Diseases, Department of Internal Medicine, Kyungpook National University Hospital, 130 Dongdoek-ro Jung-gu, Daegu 700-721, Republic of Korea. Tel.: (82) 53200-6525; fax: (82) 53-426-2046. E-mail address: [email protected] (S.W. Kim)
2
Y.K. Yoon et al. / The Spine Journal
-
(2015)
-
analysis using the presence or absence of risk factors was used to generate a risk index to identify patients with increased probability of tuberculous spondylodiscitis. RESULTS: Of 177 patients, multivariate logistic regression analysis showed that patients with tuberculous spondylodiscitis (n560) were more frequently women, with increased nonlumbar spinal involvement and associated non-spinal lesions, delayed diagnosis, higher serum albumin levels, reduced white blood cell counts, and lower C-reactive protein and procalcitonin levels. Among 117 patients with pyogenic spondylodiscitis, the most frequent causative microorganism was Staphylococcus aureus (64.1%). The mean diagnostic delay was significantly shorter, which may reflect higher clinical expression leading to earlier diagnosis. A combination of clinical data and biomarkers had better predictive value for differential diagnosis compared with biomarkers alone, with an area under the curve of 0.93, and sensitivity, specificity, and positive and negative predictive values of 95.0%, 79.5%, 70.4%, and 96.9%, respectively. CONCLUSIONS: This study provides guidance for clinicians to predict the causative organisms of spondylodiscitis in uncertain situations and before culture or pathologic examinations. Clinical data and single biomarkers combined can be useful for differential diagnoses between tuberculous and pyogenic spondylodiscitis. Ó 2015 Published by Elsevier Inc. Keywords:
Pyogenic; Tuberculosis; Spondylitis; Discitis; Leukocytes; Procalcitonin; C-reactive protein
Introduction Spondylodiscitis encompasses a wide range of vertebral osteomyelitis, spondylitis, and discitis; its increasing incidence is likely because of both an increasingly susceptible population and availability of advanced diagnostic tools [1–3]. The most severe spondylodiscitis complications include death and severe functional sequelae in 0% to 11% [4–10] and 5% to 35% [11–14] of patients, respectively. Delayed diagnosis and treatment is an independent risk factor for these adverse outcomes [7]. Because a variety of infectious pathogens can cause spondylodiscitis, it is difficult to determine its etiology without culture or pathologic examination. Therefore, specific microbiological diagnosis is preferred to confirm the infectious etiology and allow tailored antimicrobial therapy. However, the frequency of performing diagnostic biopsies varies (48%–100%) among spondylodiscitis studies [15]. Biopsy cultures from these series were positive in 43% to 90% of cases [5,7,16–19]. Therefore, diagnosis is difficult and often based on clinical, radiological, and laboratory features rather than microbiological tests. Knowledge of spondylodiscitis epidemiology is important to identify causative microorganisms globally, but this information has been very limited until recently. Most spinal infections are tuberculous or pyogenic. Previous studies in developed countries have reported tuberculous and pyogenic spondylodiscitis frequencies from 17% to 39% and 50% to 85%, respectively, among patients with spontaneous spinal infections [6,10,17,18,20–22]. Although differential diagnosis between tuberculous and pyogenic spondylodiscitis is essential for appropriate management, comparative studies of clinical, radiological, and laboratory differences to determine proper initial empirical antimicrobial therapies are limited [6,17,18,20]. The purpose of this multicenter comparative study was to describe the clinical characteristics of tuberculous and
pyogenic spondylodiscitis in the Republic of Korea, with particular focus on their differential diagnosis owing to the high tuberculosis burden. Materials and methods Study design and patients A multicenter retrospective comparative study was conducted in 11 teaching hospitals in the Republic of Korea from January 2011 to December 2013. Data were collected from the medical records of adult patients ($18 years) diagnosed with tuberculous or pyogenic spontaneous spondylodiscitis who were followed from treatment until death or end of treatment. This study included only patients diagnosed with spontaneous spondylodiscitis, except the patients with iatrogenic spondylodiscitis. Iatrogenic spondylodiscits was defined as spinal infection as a result of a penetrating injury or a surgical or radiological procedure. Only the first spondylodiscitis episodes were included in the analysis. The study protocol was approved before study initiation by the institutional review boards of each participating hospital. Definitions Pyogenic spondylodiscitis was diagnosed based on causative organisms isolated from spinal or paraspinal tissues and compatible symptoms, signs, or radiologic evidence of spondylodiscitis. Compatible spondylodiscitis symptoms or signs included spinal pain, limited motion, and tenderness over the involved spine, fever, or neurologic manifestations. Tuberculous spondylodiscitis was diagnosed when Mycobacterium tuberculosis was isolated from spinal or paraspinal tissues, smears were positive for acid-fast bacilli, tissue examination revealed caseating granuloma formation, or tissue samples were positive for
Y.K. Yoon et al. / The Spine Journal
M. tuberculosis on polymerase chain reaction (PCR), accompanied by compatible symptoms, signs, or radiologic evidence of spondylodiscitis. Candidate predictive factors were selected to guide the differential diagnosis between tuberculous and pyogenic spondylodiscitis based on the previous studies [6,17,20,23]. A relapse was defined as the reoccurrence of symptoms not attributable to other causes or new vertebral lesions on imaging together with elevated C-reactive protein (CRP) levels, increased erythrocyte sedimentation rates (ESRs), or both, after completing treatment. Diagnostic intervention to confirm the constancy of microbial diagnosis was not performed in the relapsed cases.
-
(2015)
-
3
CRP) and PCT levels that produce the best sensitivity and specificity based on the Youden index are not necessarily optimal in multivariate logistic regression analysis; therefore, a threshold analysis was performed to examine all possible combinations of cutoff values. Two-tailed p values less than .05 were considered statistically significant. All analyses were performed using IBM SPSS Statistics, version 20.0 (IBM Corporation, Armonk, NY, USA), R, version 2.15.2 (The R Foundation for Statistical Computing, Vienna, Austria), and SAS, version 9.2 (SAS Institute, Inc., Cary, NC, USA).
Results Variables
Patient characteristics
Data collected for this analysis included demographic characteristics, comorbidities, Charlson comorbidity index, [24] clinical manifestations, predisposing infection factors, time elapsed to diagnosis of spondylodiscitis, microbiological data, laboratory and radiologic data at onset, antimicrobial therapy, hospital stay, and inhospital mortality.
During the study period, data from 177 patients who met the inclusion criteria were collected from the participating hospitals. Patients were diagnosed with tuberculous spondylodiscitis (n560) or pyogenic spondylodiscitis (n5117). The same organism was identified in both tissue and blood cultures in 31 of 117 patients with pyogenic spondylodiscitis. Among 60 patients with tuberculous spondylodiscitis, M. tuberculosis was isolated from tissue samples from 50 patients; the remaining 10 patients were diagnosed based only on positive M. tuberculosis PCR results. Demographic and baseline characteristics of the 177 patients are listed in Table 1. Of these, 89 patients (50.3%) were women. The median age was 69 (IQR, 55–76) years. The proportion of women was significantly higher among patients with tuberculous spondylodiscitis compared with those with pyogenic spondylodiscitis. Univariate analyses revealed no significant age differences between the spondylodiscitis-positive groups (Table 1). The median Charlson comorbidity index was 2 (IQR, 0–2), and univariate analyses revealed that patients with pyogenic spondylodiscitis had significantly higher indices than those with tuberculous spondylodiscitis. Underlying cardiovascular diseases including hypertension, ischemic heart diseases, heart failure, and arrhythmia were significantly more common in patients with pyogenic spondylodiscitis (Table 1).
Statistical analyses Categorical and continuous variables were expressed as count (proportion) and median (interquartile range [IQR]), respectively. The Pearson chi-square test was used to compare groups of categorical variables. The Fisher exact test was used for expected case frequencies of five or less for any group. A two-sample Student t test was used to compare groups of normally distributed continuous variables. The Mann-Whitney U test was used to evaluate non-normal variable distributions. Receiver-operating characteristic (ROC) curves for each biomarker were derived and evaluated by comparing the area under the curves for assessing the accuracy of discriminable predictability between tuberculous and pyogenic spondylodiscitis, sensitivity, specificity, negative predictive value, and positive predictive value. Finally, the ROC curves of white blood cells (WBCs), CRP, and procalcitonin (PCT) were compared. Multivariate logistic regression analysis using the backward stepwise variable selection based on the Wald statistic was used to identify risk factors associated with tuberculous spondylodiscitis. Independent variables for multivariate logistic regression analysis were evaluated if they had at least 10% significance as predictors of tuberculous spondylodiscitis in the univariate analysis. The models were evaluated using the Hosmer-Lemeshow goodness-of-fit test. The predictive accuracy obtained by leave-one-out crossvalidation was used to evaluate the performance of a final logistic regression model. Receiver-operating curves were also constructed for the multivariate logistic regression model to evaluate discriminable predictability between tuberculous and pyogenic spondylodiscitis. The optimal cutoff values for individual biomarkers (i.e., WBC counts and
Clinical characteristics and predisposing factors The median diagnostic delay was 14 (IQR, 4–33) days and longer in patients with tuberculous spondylodiscitis (Table 2). Fever exceeding 38.0 C was present in 73 (62.4%) and 17 (28.3%) patients with pyogenic and with tuberculous spondylodiscitis, respectively (p!.001). There were no statistically significant differences in frequency of back pain and neurological symptoms between groups. Twenty-nine patients (16.4%) had accompanying meningitis, without significant difference in frequency between groups (Table 2).
4
Y.K. Yoon et al. / The Spine Journal
-
(2015)
-
Table 1 Comparison of demographic and baseline characteristics between patients with pyogenic or tuberculous spondylodiscitis Spondylodiscitis groups Variables*
All (n5177)
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
p Valuey
Female, n (%) Median age (y), median (IQR) Medical admission, n (%) Comorbidity, n (%) Cardiovascular disorders Central nervous system disorders Malignancy Renal disorders Hepatic disorders Respiratory disorders Connective tissue diseases Metabolic diseases Hematologic diseases Charlson comorbidity index, median (IQR)
89 (50.3) 69 (55–76) 93 (52.5)
71 (60.7) 68 (60–75) 71 (60.7)
18 (30.0) 69 (54–76) 22 (36.7)
!.001 .685 .002
61 17 18 9 15 12 5 45 7 0
47 12 15 7 31 7 4 35 5 1
14 5 3 2 2 5 1 10 2 0
(34.5) (9.6) (10.2) (5.1) (8.5) (6.8) (2.8) (25.4) (4.0) (0–2)
(40.2) (10.3) (12.8) (6.0) (11.1) (6.0) (3.4) (29.9) (4.3) (0–2)
(23.3) (8.3) (5.0) (3.3) (3.3) (8.3) (1.7) (16.7) (3.3) (0–1)
.026 .681 .103 .720 .079 .544 .663 .055 1.000 .005
IQR, interquartile range. * Values represent the number of subjects (%), mean (standard deviation), or median (IQR). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
Table 3 shows predisposing factors and non–spinal-associated lesions in patients with spontaneous spondylodiscitis. Non–spinal-associated lesions were more frequently associated with patients with tuberculous spondylodiscitis than those with pyogenic spondylodiscitis. Radiologic findings and laboratory data The most frequently involved region in both groups was the lumbar spine. However, nonlumbar involvement was significantly more frequent in patients with tuberculous spondylodiscitis (24 [40.0%] vs. 27 [23.1%], p5.019). Forty-three (24.3%) patients had involvement of three or more levels. Epidural or paraspinal abscess was present in 72 (61.5%) and 38 (63.3%) patients with pyogenic and tuberculous spondylodiscitis, respectively (p5.816). Table 2 summarizes laboratory findings in both groups. Compared with those with tuberculous spondylodiscitis, patients with pyogenic spondylodiscitis were more frequently associated with the following: hematocrit less than 30%, WBC count greater than or equal to 15,000/mm3, proportion of neutrophils greater than 75%, serum albumin level less than 3.0 g/dL, and CRP level greater than 7.25 mg/L (Table 2). Fig. 1 represents the differential diagnostic accuracy of the biomarkers for patients with spontaneous spondylodiscitis, estimated with the area under the curve (AUC) according to the cutoff values of the single biomarkers. Microbiological characteristics All 177 patients in this study underwent tissue examination. Among 117 patients with pyogenic spondylodiscitis, 2 had polymicrobial samples, with methicillin-resistant Staphylococcus aureus (MRSA), Streptococcus agalactiae, methicillin-susceptible S. aureus (MSSA), and Pseudomonas
aeruginosa. The other 115 pyogenic spondylodiscitis cases were monomicrobial, with the following isolated organisms: MSSA (n555), MRSA (n518), coagulase-negative Staphylococcus (n510), Streptococcus spp. (n511), Enterococcus faecalis (n51), Escherichia coli (n59), Klebsiella pneumoniae (n56), P. aeruginosa (n51), Serratia marcescens (n51), Citrobacter freundii (n51), Proteus mirabilis (n51), and Salmonella Group C (n51). Among 60 patients with tuberculous spondylodiscitis, M. tuberculosis was isolated from tissues in 50 patients, and 43 were for positive for M. tuberculosis on PCR. Treatment outcomes Table 4 shows treatment and clinical outcomes. The overall all-cause inhospital and spondylodiscitis-related mortalities were 5.6% (10/177) and 2.8% (5/177), respectively. There were no significant differences in all-cause inhospital and spondylodiscitis-related mortality between groups. The frequency of surgical intervention because of spondylodiscitis was higher among patients with tuberculous spondylodiscitis compared with those with pyogenic spondylodiscitis. After diagnosis of spontaneous spondylodiscitis, the median hospital stay was significantly longer in patients with pyogenic spondylodiscitis. Factors for differential diagnosis between groups Multiple logistic regression modeling revealed that the statistically significant factors associated with tuberculous spondylodiscitis included sex (female, odds ratio, 8.84; 95% confidence interval, 2.78–28.07), involvement of nonlumbar spine (3.48, 1.16–10.47), elapsed to time spondylodiscitis diagnosis more than 7 days (4.79, 1.58–14.51), WBC count less than or equal to 9,600/mm3 (29.82, 6.34– 140.17), a combination of serum albumin levels greater than
Y.K. Yoon et al. / The Spine Journal
-
(2015)
-
5
Table 2 Comparison of clinical characteristics between patients with pyogenic or tuberculous spondylodiscitis Spondylodiscitis groups All (n5177)
Variables* Clinical features, n (%) Time elapsed to admission (d), median (IQR) Time elapsed to diagnosis of spondylodiscitis #7 d Fever Back pain Neurological symptoms Meningitis, n (%) Involved vertebral regions, n (%) Cervical Thoracic Lumbar Sacral $3 spinal regions Skipping lesions Paraspinal abscess Epidural abscess Psoas muscle abscess Laboratory results at time of initial diagnosis, n (%) Anemia (hematocrit !30) Leukocyte count $15,000/mm3 Neutrophilia $75% Hypoalbuminemia !3.0 g/dL WBC count (per mm3), median (IQR) CRP level (mg/L), median (IQR) CRP level O7.25 mg/L ESR (mm/h), median (IQR) ALP level (IU/L), median (IQR) PCT level (ng/dL), median (IQR) Interferon-gamma release assays
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
14 118 90 149 77 29
(4–30) (66.7) (50.8) (84.2) (43.5) (16.4)
10 67 73 102 48 10
(4–29) (57.3) (62.4) (87.2) (41.0) (16.7)
30 51 17 47 29 19
(12–73) (85.0) (28.3) (78.3) (48.3) (16.2)
17 48 126 19 43 6 62 87 43
(9.6) (27.1) (71.2) (10.7) (24.3) (3.4) (35.0) (49.2) (24.3)
10 24 90 15 25 4 38 58 29
(8.5) (20.5) (76.9) (12.8) (21.4) (3.4) (32.5) (49.6) (24.8)
7 24 36 4 18 2 24 29 14
(11.7) (40.0) (60.0) (6.7) (30.0) (3.3) (40.0) (48.3) (23.3)
55 30 113 48 9,300 12.5 107 63 111 0.51 22/42
(31.1) (16.9) (63.8) (27.1) (6,970–13,450) (3.3–34.7) (60.5) (40–83) (60–216) (0.082–1.41) (52.4)
43 30 89 44 11,900 19.9 88 63 129.0 0.59 5/18
(36.8) (25.6) (76.1) (37.6) (7,650–15,560) (6.9–80.7) (75.2) (39–86) (68.8–217.0) (0.11–1.67) (27.8)
12 (20.0) 0 24 (40.0) 4 (6.7) 7,450 (5,655–8,882) 3.2 (1.8–12.3) 19 (31.7) 62 (40–80) 94 (31–205) 0.11 (0.06–0.70) 17/24(70.8)
p Valuey !.001 !.001 !.001 .127 .353 .942 .505 .006 .019 .211 .266 1.000 .321 .876 .831 .023 !.001 !.001 !.001 !.001 !.001 !.001 .519 .098 .066 .006
ALP, alkaline phosphatase; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; PCT, procalcitonin; WBC, white blood cell. * Values represent the number of subjects (%), mean (standard deviation), or median (interquartile range). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
or equal to 3.0 mg/dL and PCT levels less than or equal to 1.39 ng/dL (8.97, 1.65–48.69), and a combination of non– spinal-associated lesions and CRP levels #7.25 mg/L (8.59, 1.62–45.64) (Table 5). The p value for the HosmerLemeshow goodness-of-fit test was 0.954, greater than the 0.05 significance threshold; therefore, there is no significant evidence for lack of fit in any of the final models.
Leave-one-out cross-validation was performed to assess the predictive accuracy of each final model. The AUCs for the clinical failure model were about 0.90 for both raw data and leave-one-out cross-validation. For this model, the sensitivity, specificity, and positive predictive value and negative predictive value obtained with an optimal cutoff point were greater than 0.70 (Fig. 2).
Table 3 Comparison of predisposing factors between patients with pyogenic or tuberculous spondylodiscitis Spondylodiscitis groups Variables*
All (n5177)
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
p Valuey
Invasive spinal procedures Previous admission Previous operations Recent bacteremia Exposure to immunosuppressant History of tuberculosis Non–spinal-associated lesions
7 61 16 30 7 29 56
4 42 11 22 4 23 31
3 19 5 8 3 6 25
.690 .575 .814 .359 .690 .100 .040
(4.0) (34.5) (9.0) (16.9) (4.0) (16.4) (31.6)
(3.4) (35.9) (9.4) (18.8) (3.4) (19.7) (26.5)
(5.0) (31.7) (8.3) (13.3) (5.0) (10.0) (41.7)
* Values represent the number of subjects (%), mean (standard deviation), or median (interquartile range). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
6
Y.K. Yoon et al. / The Spine Journal
Fig. 1. Receiver-operating characteristic curve of white blood cell counts, C-reactive protein levels, and procalcitonin levels to differentiate patients with tuberculous or pyogenic spondylodiscitis.
Discussion This multicenter comparative study investigated the clinical, laboratory, and radiological characteristics of spontaneous spondylodiscitis secondary to tuberculosis and other bacterial causes. In this study, predictors useful for differential diagnosis between the spondylodiscitis groups before culture or pathologic examinations were identified, and ROC curve analysis using serum WBC counts and CRP and PCT levels produced a risk index to identify patients with an increased probability of tuberculous spondylodiscitis. Our findings suggest that a combination of information derived from clinical examination and laboratory parameters, such as WBC counts and CRP or PCT levels, may be useful for differential diagnosis between tuberculous and pyogenic spondylodiscitis. Tuberculosis is a global concern in developing and developed countries. Despite rapid economic development, its prevalence in the Republic of Korea is the highest among Organization for Economic Cooperation and Development nations, with an incidence of 108 cases per 100,000 people [25]. Patients with spinal tuberculosis are not
-
(2015)
-
common but comprise up to 50% of all patients afflicted with musculoskeletal tuberculosis. In tuberculosisendemic countries such as the Republic of Korea, a clinical suspicion of tuberculous spondylodiscitis is required for early selection of appropriate antimicrobial therapies. Data to guide the differential diagnosis between tuberculous and pyogenic spondylodiscitis have been identified in a few single-center studies with small sample sizes. Known risk factors for tuberculous spondylodiscitis included chronic renal failure, absence of fever, constitutional symptoms, neurological deficit, psoas abscess, history of tuberculosis, active tuberculosis of other organs, increased diagnostic delay, presence of spinal deformities, involvement of the thoracic spine or greater than or equal to three spinal levels, higher ESR, and involvement of posterior elements [6,17,20]. Risk factors for pyogenic spondylodiscitis included previous invasive spinal procedures, preceding bacteremia, diabetes mellitus, underlying chronic debilitating diseases, chronic renal failure, liver cirrhosis, intravenous drug abuse, immunosuppression, fever (temperature O38 C), WBC count greater than 1,000/mm3, neutrophil fractions greater than 75%, CRP level greater than 5 mg/ dL, ESR greater than 40 mm/h, and alkaline phosphatase level greater than 120 IU/L [6,20]. Several predictors identified in our study were comparable with the previous reports [17,20]. The mean diagnostic delay was significantly shorter among patients with pyogenic spondylodiscitis, which may reflect higher clinical expression that permits an earlier diagnosis. Although tuberculosis affects men more often than women, women were more frequently diagnosed with tuberculous spondylodiscitis, another finding similar to the other studies [20,25]. Previous studies reported lumbar and thoracic spines to be the most common regions of spondylodiscitis among patients with pyogenic and tuberculous spondylodiscitis, respectively [23]. However, we found the lumbar region to be most commonly affected region in both groups. Involvement of the thoracic or cervical spine was more frequently associated with tuberculous spondylodiscitis (Table 2). In this study, the 5.6% mortality rate among patients with pyogenic spondylodiscitis was similar to previously
Table 4 Comparison of treatment and clinical outcomes among patients with pyogenic and tuberculous spondylodiscitis Spondylodiscitis groups Variables* Antimicrobial treatment Duration of appropriate treatment (d), median (IQR) Operation because of spondylitis Outcomes, n (%) Treatment relapse Complications Hospital stay after onset of spondylitis (d), median (IQR) Inhospital mortality, n (%) Mortality attributable to spondylitis, n (%)
All (n5177)
Pyogenic (n5117, 66.1%)
Tuberculous (n560, 33.9%)
p Valuey
76 (30–312) 102 (57.6)
56 (22–89) 58 (49.6)
360 (257–368) 44 (73.3)
!.001 .002
9 22 41 10 5
(5.1) (12.4) (26–65) (5.6) (2.8)
7 12 47 8 5
(6.0) (10.3) (34–67) (6.8) (4.3)
IQR, interquartile range. * Values represent the number of subjects (%), mean (standard deviation), or median (IQR). y p Values were obtained using Student t, Mann-Whitney U, Fisher exact, or chi-square test as appropriate.
2 (3.3) 10 (16.7) 30 (17–55) 2 (3.3) 0
.720 .221 .001 .498 .168
Y.K. Yoon et al. / The Spine Journal Table 5 Multivariate logistic regression analyses of risk factors associated with tuberculous spondylodiscitis in 177 patients with pyogenic or tuberculous spondylodiscitis Independent variables* Female sex Involvement of nonlumbar spine Time elapsed to diagnosis of spondylodiscitis O7 d WBC count #9,600/mm3 Serum albumin level $3.0 mg/dL and PCT level #1.39 ng/dL Associated non-spinal lesions and CRP level #7.25 mg/L
OR (95% CI for OR)
p Value
8.84 (2.78–28.07) 3.48 (1.16–10.47) 4.79 (1.58–14.51)
!.001 .027 .006
29.82 (6.34–140.17) 8.97 (1.65–48.69)
!.001 .011
8.59 (1.62–45.64)
.012
CI, confidence interval; CRP, C-reactive protein; OR, odds ratio; PCT, procalcitonin; WBC, white blood cell. * Female sex, diabetes mellitus, Charlson comorbidity index, fever ($38 C), involvement of thoracic spine, involvement of nonlumbar spine, non–spinal-associated lesions, time elapsed to diagnosis of spondylodiscitis O7 days, hematocrit !30%, serum albumin $3.0 mg/dL, WBC count #9,600/mm3, CRP level #7.25 mg/L, and PCT level #1.39 ng/dL were evaluated as independent variables for multivariable logistic regression analysis if they were predictors of tuberculous spondylodiscitis at the 10% significance level.
reported rates of 4% to 10% [18,26,27]. No patients with tuberculous spondylodiscitis died, comparable with mortality rates of 0% and 7.3% reported by the previous studies and attributed to other causes [6,20]. Spondylodiscitis prognosis might depend on early diagnosis and treatment or identification of the causal agent [18,26,27]. In our study, the five mortality cases attributable to spondylodiscitis were pyogenic spondylodiscitis, caused by MRSA (n53), MSSA (n51), and Streptococcus viridans (n51), with an elapsed time to diagnosis of 14 days or more, except for a single patient. Similar to the previous reports, surgical
-
(2015)
-
7
treatment was more frequently associated with tuberculous than pyogenic spondylodiscitis [17]. Our study also found that WBC counts and CRP and PCT levels might be valuable laboratory findings to diagnose spondylodiscitis. Univariate analyses indicated that higher levels of these biomarkers were suggestive of pyogenic rather than tuberculous spondylodiscitis. The AUC for these three biomarkers was approximately 0.7 (Fig. 1). Multivariate logistic regression modeling showed that clinical information along with laboratory findings resulted in better differential diagnosis between spondylodiscitis groups compared with the discriminative ability of each individual biomarker (Fig. 2). This study has some limitations. First, this study was performed in a tuberculosis-endemic region; therefore, these results may not be generalizable to other populations or areas with low tuberculosis rates. Second, all subjects in this study were microbiologically confirmed patients. However, the causative microorganisms in clinical practice are often not identified microbiologically in patients with spondylodisictis [1,28]. The predictability of these risk factors to discriminate between patients with tuberculous and pyogenic spondylodiscitis was not evaluated in those with a negative culture. Further research is required to determine whether these predictive factors may be more helpful for deciding initial empirical antibiotics before the identification of the causative microorganism or for planning the antibiotic regimen in the patients who were not microbiologically confirmed. Third, the ranges in the quantitative indices of the laboratory biomarkers used in this study had a wide overlap between pyogenic and tuberculous spondylodiscitis. Therefore, it was difficult to determine an accurate cutoff value for discriminating each biomarker. However, the cutoff values in the multivariate logistic regression analyses were determined at the highest AUC value in conjunction with other predictive factors (Table 5). In conclusion, this multicenter comparative study suggests that a combination of clinical data and laboratory findings can be useful to discriminate between patients with tuberculous and pyogenic spondylodiscitis. Particularly, these predictive factors may be useful for selecting the initial empirical antibiotics before culture results are available. Further prospective randomized trials with larger study populations are necessary to develop better predictive models to guide selection of appropriate initial empirical antimicrobial therapies in patients with spondylodiscitis. Acknowledgments
Fig. 2. Receiver-operating characteristic curve for tuberculous spondylodiscitis obtained using the predictive probability of multivariate logistic regression model and validation results.
This work was supported in part by the Korean Society for Chemotherapy and by Roche, in the Republic of Korea. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Institutional review board approval/research ethics committee: The study protocol was approved before study
8
Y.K. Yoon et al. / The Spine Journal
initiation by the institutional review boards of each participating hospital. References [1] Lora-Tamayo J, Euba G, Narvaez JA, Murillo O, Verdaguer R, Sobrino B, et al. Changing trends in the epidemiology of pyogenic vertebral osteomyelitis: the impact of cases with no microbiologic diagnosis. Semin Arthritis Rheum 2011;41:247–55. [2] Cottle L, Riordan T. Infectious spondylodiscitis. J Infect 2008;56: 401–12. [3] Lew DP, Waldvogel FA. Osteomyelitis. N Engl J Med 1997;336: 999–1007. [4] Legrand E, Flipo RM, Guggenbuhl P, Masson C, Maillefert JF, Soubrier M, et al. Management of nontuberculous infectious discitis. Treatments used in 110 patients admitted to 12 teaching hospitals in France. Joint Bone Spine 2001;68:504–9. [5] Beronius M, Bergman B, Andersson R. Vertebral osteomyelitis in G€ oteborg, Sweden: a retrospective study of patients during 199095. Scand J Infect Dis 2001;33:527–32. [6] Colmenero JD, Jimenez-Mejıas ME, Sanchez-Lora FJ, Reguera JM, Palomino-Nicas J, Martos F, et al. Pyogenic, tuberculous, and brucellar vertebral osteomyelitis: a descriptive and comparative study of 219 cases. Ann Rheum Dis 1997;56:709–15. [7] McHenry MC, Easley KA, Locker GA. Vertebral osteomyelitis: longterm outcome for 253 patients from 7 Cleveland-area hospitals. Clin Infect Dis 2002;34:1342–50. [8] Friedman JA, Maher CO, Quast LM, McClelland RL, Ebersold MJ. Spontaneous disc space infections in adults. Surg Neurol 2002;57:81–6. [9] Wirtz DC, Genius I, Wildberger JE, Adam G, Zilkens KW, Niethard FU. Diagnostic and therapeutic management of lumbar and thoracic spondylodiscitis—an evaluation of 59 cases. Arch Orthop Trauma Surg 2000;120:245–51. [10] Nolla JM, Ariza J, Gomez-Vaquero C, Fiter J, Bermejo J, Valverde J, et al. Spontaneous pyogenic vertebral osteomyelitis in nondrug users. Semin Arthritis Rheum 2002;31:271–8. [11] Ross PM, Fleming JL. Vertebral body osteomyelitis: spectrum and natural history. A retrospective analysis of 37 cases. Clin Orthop Relat Res 1976;118:190–8. [12] Osenbach RK, Hitchon PW, Menezes AH. Diagnosis and management of pyogenic vertebral osteomyelitis in adults. Surg Neurol 1990;33:266–75. [13] Griffiths HE, Jones DM. Pyogenic infection of the spine. A review of twenty-eight cases. J Bone Joint Surg Br 1971;53:383–91.
-
(2015)
-
[14] Cahill DW, Love LC, Rechtine GR. Pyogenic osteomyelitis of the spine in the elderly. J Neurosurg 1991;74:878–86. [15] Mylona E, Samarkos M, Kakalou E, Fanourgiakis P, Skoutelis A. Pyogenic vertebral osteomyelitis: a systematic review of clinical characteristics. Semin Arthritis Rheum 2009;39:10–7. [16] Hadjipavlou AG, Mader JT, Necessary JT, Muffoletto AJ. Hematogenous pyogenic spinal infections and their surgical management. Spine 2000;25:1668–79. [17] Turunc T, Demiroglu YZ, Uncu H, Colakoglu S, Arslan H. A comparative analysis of tuberculous, brucellar and pyogenic spontaneous spondylodiscitis patients. J Infect 2007;55:158–63. [18] Perronne C, Saba J, Behloul Z, Salmon-Ceron D, Leport C, Vilde JL. Pyogenic and tuberculous spondylodiscitis (vertebral osteomyelitis) in 80 adult patients. Clin Infect Dis 1994;19:746–50. [19] Karadimas EJ, Bunger C, Lindblad BE, Hansen ES, Høy K, Helmig P. Spondylodiscitis. A retrospective study of 163 patients. Acta Orthop 2008;79:650–9. [20] Kim CJ, Song KH, Jeon JH, Park WB, Park SW, Kim HB, et al. A comparative study of pyogenic and tuberculous spondylodiscitis. Spine 2010;35:E1096–100. [21] Luzzati R, Giacomazzi D, Danzi MC, Tacconi L, Concia E, Vento S. Diagnosis, management and outcome of clinically-suspected spinal infection. J Infect 2009;58:259–65. [22] Belzunegui J, Del Val N, Intxausti JJ, De Dios JR, Queiro R, Gonzalez C, et al. Vertebral osteomyelitis in northern Spain. Report of 62 cases. Clin Exp Rheumatol 1999;17:447–52. [23] Lee KY. Comparison of pyogenic spondylitis and tuberculous spondylitis. Asian Spine J 2014;8:216–23. [24] Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol 1994;47:1245–51. [25] World Health Organization. Global tuberculosis control. Geneva, Switzerland: World Health Organization, 2013. [26] Buchelt M, Lack W, Kutschera HP, Katterschafka T, Kiss H, Schneider B, et al. Comparison of tuberculous and pyogenic spondylitis. An analysis of 122 cases. Clin Orthop Relat Res 1993;296: 192–9. [27] Kim J, Kim YS, Peck KR, Kim ES, Cho SY, Ha YE, et al. Outcome of culture-negative pyogenic vertebral osteomyelitis: comparison with microbiologically confirmed pyogenic vertebral osteomyelitis. Semin Arthritis Rheum 2014;44:246–52. [28] Bhagat S, Mathieson C, Jandhyala R, Johnston R. Spondylodiscitis (disc space infection) associated with negative microbiological tests: comparison of outcome of suspected disc space infections to documented non-tuberculous pyogenic discitis. Br J Neurosurg 2007;21: 473–7.
