Clinical Characteristics and Risk Factors of Pyogenic Spondylitis Caused by Gram-Negative Bacteria

Background There are limited data describing the clinical characteristics of pyogenic spondylitis caused by Gram-negative bacteria (GNB). The aim of this study was to investigate the predisposing factors and clinical characteristics of pyogenic spondylitis caused by GNB compared to Gram-positive cocci (GPC). Methods We performed a retrospective review of medical records from patients with culture-confirmed pyogenic spondylitis at four tertiary teaching hospitals over an 8-year period. Results A total of 344 patients with culture-confirmed pyogenic spondylitis were evaluated. There were 62 patients (18.0%) with pyogenic spondylitis caused by GNB and the most common organism was Escherichia coli (n = 35, 10.2%), followed by Pseudomonas aeruginosa (n = 10, 2.9%). Pyogenic spondylitis caused by GNB was more frequently associated with the female gender (64.5 vs. 35.5%, P <0.01), preexisting or synchronous genitourinary tract infection (32.3 vs. 2.1%, P< 0.01), and intra-abdominal infection (12.9 vs. 0.4%, P< 0.01) compared to patients with GPC. Although pyogenic spondylitis caused by GNB presented with severe sepsis more frequently (24.2 vs. 11.3%, P = 0.01), the mortality rate (6.0 vs. 5.2%) and the proportion of patients with residual disability (6.0 vs. 9.0%), defined as grade 3 or 4 (P = 0.78) 3 months after completion of treatment, were not significantly different compared to GPC patients. Conclusion GNB should be considered as the etiologic organism when infectious spondylitis develops in a patient with preexisting or synchronous genitourinary tract and intra-abdominal infection. In addition, the mortality rate and clinical outcomes are not significantly different between pyogenic spondylitis caused by GNB and GPC.


Introduction
Infectious spondylitis is a destructive infection of the spine or paraspinal structures [1,2]. As spondylitis progresses, vertebrae are destroyed, and inflammation extends to the epidural and paraspinal spaces. The most detrimental consequence of infectious spondylitis is neurologic deficits, which develop in approximately one-third of cases.
Microbiologic diagnosis is an important step in the management of pyogenic spondylitis, because it requires long-term antibiotic treatment. However, despite various efforts, in some patients, etiologic organisms are not isolated and antibiotics are chosen empirically. The most commonly used empirical antibiotics are first-generation cephalosporins, which are active against methicillin-susceptible S. aureus and streptococci [8]. The antibiotic susceptibilities of GNB vary between species, and are frequently not susceptible to first-generation cephalosporins. Predictive factors for GNB in patients with pyogenic spondylitis are critical for choosing empirical antibiotics. However, few studies have identified the predisposing factors or clinical characteristics of pyogenic spondylitis caused by GNB. Moreover, studies of GNB pyogenic spondylitis report results from less than 10 patients, with the exception of one study that reported data from 65 patients [9].
The objectives of this study were to investigate the predisposing factors for GNB in patients with pyogenic spondylitis and the clinical characteristics of pyogenic spondylitis caused by GNB.

Study design
We conducted a retrospective cohort study at four university-affiliated teaching hospitals from January 2005-March 2013. We retrospectively collected medical records of patients 18 years of age with pyogenic spondylitis. Only culture-confirmed patients with pyogenic spondylitis were included, and patients with infectious spondylitis caused by Brucella species, M. tuberculosis, or fungi polymicrobial infection were excluded. We collected demographic data, co-morbidities, presenting symptoms, microbiologic data, radiographic characteristics, and laboratory data including white blood cell (WBC) counts, C-reactive protein (CRP), surgical treatment, clinical outcomes, and mortality rate. The presence of the following predisposing factors was also documented: preexisting or synchronous infection, presence of spinal prosthesis, spinal surgery within 1 year, and previous invasive spinal procedures within 6 months, such as epidural block, acupuncture and percutaneous vertebroplasty. Clinical outcomes of mortality and neurologic outcome were evaluated at 3 months and 12 months after completion of treatment. To investigate predisposing factors and clinical characteristics, we compared GNB pyogenic spondylitis with Gram-positive cocci (GPC) spondylitis. The Institutional Review Boards of the four participating hospitals (Chonnam National University Hospital Institutional Review Board, Chonnam National University Hwasun Hospital Institutional Review Board, Hwasun, Seoul National University Hospital Institutional Review Board, and the Seoul National University Bundang Hospital Institutional Review Board) approved this study. A waiver of consent was granted, given the retrospective nature of the project.

Definition
Pyogenic spondylitis was diagnosed when the etiologic organism was isolated from spinal or paraspinal tissues, or if there were compatible clinical signs or symptoms and radiologic evidence of vertebral infection, and positive blood cultures which were performed at the time of diagnosis of pyogenic spondylitis, including at least two separate sets of blood cultures in the case of common skin contaminants, such as coagulase-negative staphylococci. Compatible signs or symptoms of vertebral infection were defined as pain, fever, or neurologic manifestations. Typical MRI characteristics included decreased signal intensity in the vertebral body and disk and loss of end plate on T1-weighted images and increased signal intensity of the disk and vertebral body on T2-weighted images [10]. Preceding bacteremia was defined as bacteremia from any bacteria within 12 months before the diagnosis of spondylitis [4].
Preexisting or synchronous infection was defined as a documented infection at another site within the 30 days prior to or at the diagnosis of pyogenic spondylitis [10]. Severe sepsis was defined as sepsis with one or more signs of organ dysfunction. Neurologic staging was divided into five grades, which was adopted from previous report [11]; no pain (grade 0), back pain at the level of the affected spine (grade 1), nerve-root pain radiating from involved spinal area (grade 2), motor weakness, sensory deficit, or bladder/bowel dysfunction (grade 3), and paralysis (grade 4). Outcomes were evaluated using overall mortality, duration of treatment, and residual disability. Residual disability was defined as grade 3 or 4 by neurologic staging.

Statistical analysis
The Student t-test or Mann-Whitney U-test was used to compare continuous variables. The Chi-square test or Fisher's exact test was used to compare categorical variables. A logistic regression model was adopted to adjust for confounding variables and to identify risk factors. To this end, we performed a stepwise multivariate logistic regression analysis. Variables with P < 0.10 using univariate analysis were candidates for multivariate analysis. The odds ratio (OR) and 95% confidential intervals (CIs) were calculated. All P values were two-tailed, and P < 0.05 was considered to indicate statistical significance. Statistical analysis was performed using the SPSS software (version 20.0; SPSS Inc., Chicago, IL, USA).

Antibiotic susceptibilities of GNB
The antibiotic susceptibilities of GNB causing pyogenic spondylitis are summarized in Table 2. Of 52 GNB isolates belonging to the Enterobacteriaceae family, 30.8% were resistant to fluoroquinolone, and 15.4% to third-generation cephalosporins. We identified 2 isolates from a total of 10 P. aeruginosa samples that were resistant to fluoroquinolone and 3 isolates that were resistant to anti-pseudomonal third-generation cephalosporins. The resistance rates to fluoroquinolone or third generation cephalosporins were not different in patients with or without previous spinal surgery or an invasive spinal procedure.

Clinical and radiographic characteristics of pyogenic spondylitis caused by GNB
The demographic findings and clinical characteristics of pyogenic spondylitis caused by GNB and GPC are shown in Table 3. The mean age (± SD) was similar in both groups. Female gender was more frequently associated with GNB infection (64.5 vs. 35.5%, P< 0.01). The most

Serratia marcescens
frequently reported symptom at diagnosis was pain at the affected site in each group. Fever and severe sepsis were observed more frequently in the GNB group (67.7 vs. 53.2%, P = 0.05; 24.2 vs. 11.3%, P = 0.01, respectively). The most frequently involved vertebral level was the lumbar spine in both groups, and no difference was observed in the involved vertebral level or paraspinal involvement between the two groups. The proportion of patients initially presenting with grade 3 or 4 was not different (16.4 vs. 18.2%, P = 0.85) between the GNB and GPC groups.

Predisposing factors for pyogenic spondylitis caused by GNB
A preexisting, defined as within the 30 days prior to admission, or synchronous non-spinal infection was documented in 98 (28.5%) patients and more frequently in GNB spondylitis (50.0 vs. 23.8%, P< 0.01; Table 4). Genitourinary tract infection (32.3 vs. 2.1%, P< 0.01) and intra-abdominal infection (12.9 vs. 0.4%, P< 0.01) were also more common in the GNB group. In contrast, skin and soft tissue infection, postoperative wound infection, and infective endocarditis were more common in the GPC group, albeit not significantly so. A history of preceding epidural block within 6 months was reported more frequently in the GPC group than in the GNB group (6.5 vs. 16.3%, P = 0.05). The proportion of patients who had previous invasive spinal procedures, other than epidural block, was not different between the GNB and GPC groups (Table 3). In multivariate analysis, female gender, preexisting or synchronous genitourinary tract, and intra-abdominal infection were found to be significantly associated with GNB infection (Table 5).

Clinical outcomes
The most frequent antibiotics used in patients with GNB pyogenic spondylitis were third generation cephalosporins (40.3%), followed by carbapenem (25.8%), and quinolone (21.0%). Thirty-five (56.5%) patients in the GNB group and 147 patients (52.1%) in the GPC group underwent surgical treatment (P = 0.58). The most common surgical treatment in patients with pyogenic spondylitis caused by GNB was a simple abscess drainage with irrigation (51.4%, 18/ 35), followed by neurologic complications defined by grade 3 or 4 (20%, 7/35), intractable pain    another hospital during treatment. The overall mortality rate was 6.0% and 5.2% in each group at 3 months (P = 0.73), and 6.7% and 7.3% in each group at 12 months (P>0.99). The neurologic outcome was evaluated in 283 patients 3 months after treatment and 6.0% (3/50) of patients in the GNB group and 9.0% (21/233) of patients in the GPC group had residual disability defined as grade 3 or 4 (P = 0.78).

Discussion
In the current study, 344 patients were microbiologically diagnosed with pyogenic infectious spondylitis. The proportion of patients with GNB was 18.0%, and the most common bacterium was E. coli, followed by P. aeruginosa, which is consistent with previous reports. Presenting symptoms, involvement of the spine, prevalence of paraspinal abscess, and laboratory findings were not different between the GNB and GPC groups. The most common presenting symptom of patients with GNB pyogenic spondylitis was pain at the affected site, and 16.4% of patients presented with motor weakness. The lumbar spine was the most commonly involved site, and 48.4% of patients had paraspinal abscesses. Data regarding the clinical characteristics of pyogenic spondylitis caused by GNB are scarce. All 7 patients with pyogenic spondylitis caused by E. coli and the 10 patients caused by various GNB were > 63 years of age in previous studies [6,12]. However, in this study, the age distribution of the 62 patients with GNB pyogenic spondylitis was 37-85 years. In another study, patients with GNB pyogenic spondylitis had a lower prevalence of spinal epidural abscesses, paraspinal abscesses, WBC counts, and CRP values compared with methicillin-susceptible S. aureus (MSSA) pyogenic spondylitis [9]. These results differ from our findings and could have originated from different comparative groups since the previous study compared patients with pyogenic spondylitis caused by MSSA, while in the current work, patients with GPC including MSSA were used as a comparison. Furthermore, it has been reported that MSSA has a potent ability to form abscesses [13,14]. Multivariate analysis determined that the female gender, preexisting or synchronous genitourinary tract and intra-abdominal infections were associated with GNB infection in pyogenic spondylitis. Such an association is consistent with previous reports [7,9]. For the treatment of pyogenic spondylitis, empirical antibiotics are usually not recommended unless the patient is critically ill, neutropenic, or neurologically compromised [1]. The microbiological diagnosis of these infections is particularly important because long-term antibiotic therapy is required. A microbiological diagnosis was attempted using blood or tissue cultures. However, several studies have reported negative microbiological results in up to 30% of cases [15][16][17]. In this study, microorganisms were not identified in 24.1% of cases. In culture-negative infectious spondylitis, empirical antibiotics should be selected based on the clinical context. The suspicion of GNB as a cause of pyogenic spondylitis is important because inadequate empirical antibiotic treatments for GNB could be associated with neurologic complications [6]. Our study suggests that the use of antibiotics covering GNB, such as a third-or fourth-generation cephalosporin or carbapenem, should be considered when patients are female or have had a preexisting or synchronous genitourinary tract and intra-abdominal infection. Identifying a potential source of infection, other than the spine, is crucial to the selection of empirical antibiotics to eradicate the infection. Preceding or synchronous infection at sites other than spine were reported in 30-50% of patients with pyogenic spondylitis [3,18]. The most commonly reported site of preceding infection was the genitourinary tract, followed by skin and soft tissues, respiratory tract, gastrointestinal tract, or oral cavity [11]. In our study, we identified extra-spinal infection in 28.5% of patients, and the most frequent infection site was the genitourinary tract, followed by skin and soft tissue, endocardium, intra-abdomen, and catheter. Pyogenic spondylitis caused by GNB was more frequently associated with genitourinary tract and intra-abdominal infections, and it could be surmised that GNB are the most common etiologic organisms of these infections. We determined that female gender was predominant in GNB infection, in contrast to other studies stating a fivefold male predominance in pyogenic spondylitis [19,20]. The predominance of females with GNB infection might be explained by a higher prevalence of urinary tract infection. In our study, the fluoroquinolone resistance rate in the Enterobacteriaceae family was >30%, and while not associated with a previous invasive spinal procedure, it was associated with increasing antimicrobial resistance in GNB globally [21].
When a diagnosis of pyogenic spondylitis is suspected clinically and radiologically, there is no universal recommendation for treatment. Some experts recommended a combination of levofloxacin and rifampin, or a combination of ciprofloxacin and clindamycin [1,22,23]. However, the existence of fluoroquinolone-resistant bacteria increases the risk of treatment failure and the selection of resistant mycobacteria in regions where tuberculosis remains prevalent. Further study of the risk factors associated with resistant organisms in pyogenic spondylitis is necessary.
Our study had several limitations. First, these data were collected retrospectively. Thus, unrecognized predisposing factors could have been distributed unequally between GPC and GNB pyogenic spondylitis. Second, this study included postoperative infection, as well as, healthcare-associated infection, even though the proportion of patients who had previous invasive spinal procedures, other than epidural block, was not different between the GNB and GPC groups. Therefore, the results of this study may not be truly indicative of community-onset pyogenic spondylitis.
In conclusion, female gender, preexisting or synchronous genitourinary tract, and intra-abdominal infection are predictive of GNB infection in pyogenic spondylitis.