Predicting residual neurologic deficits using the Spinal Infection Treatment Evaluation score after surgery for thoracic and lumbar spinal epidural abscess: a retrospective study in Taiwan
Article information
Abstract
Study Design
Retrospective, consecutive case series.
Purpose
We assessed the predictive validity of the Spinal Infection Treatment Evaluation (SITE) score to determine the likelihood of residual neurological deficits (ND) following surgical treatment of spinal epidural abscess (SEA).
Overview of Literature
SEA is a severe spinal infection that can result in irreversible ND and sepsis if left untreated. Although various risk factors have been proposed to predict postoperative neurological outcomes, the optimal predictors remain unclear.
Methods
A total of 45 patients diagnosed with de novo thoracic or lumbar SEA who underwent posterior-only surgical decompression between 2005 and 2014, with a minimum postoperative follow-up of 2 years, were included. Patients were stratified based on the presence or absence of postoperative residual ND, and neurological function was assessed immediately after surgery and at the final follow-up using the Frankel grading system. SITE scores, along with clinical and radiological data associated with residual ND, were collected. Logistic regression and receiver operating characteristic (ROC) curve analyses were performed to identify significant predictors.
Results
Patients with residual ND had significantly lower SITE scores than those without residual ND (4.3±1.3 vs. 7±1.8, p<0.0001). Multivariate analysis identified the SITE score as an independent predictor (odds ratio, 2.70; p=0.012). ROC analysis showed that a SITE score ≤6 predicted residual ND with 73.3% sensitivity and 100% specificity, with an area under the curve of 0.877 (p<0.001). Other significant predictors included cauda equina syndrome and a shorter symptom-to-surgery interval, both of which were associated with a higher risk of residual ND.
Conclusions
The SITE score is a reliable and independent predictor of residual ND after surgery for SEA. SITE scores <6 indicate a significantly higher risk of postoperative ND.
Introduction
Spinal epidural abscess (SEA) is a severe spinal infection characterized by the accumulation of purulent material within the epidural space. If left untreated, it can lead to irreversible neurological deficits (ND) and life-threatening sepsis in severe cases [1]. The current medical consensus recommends emergency debridement combined with systemic antibiotic therapy as the standard treatment [2,3].
The pathogenesis of postoperative ND in SEA is multifactorial and involves the direct mechanical compression of neural structures, vascular ischemia, thrombophlebitis, or a combination of these mechanisms [1,4]. Reports regarding predictors of postoperative residual ND are inconsistent, with incidence rates ranging from 11.1% to 57.1% [5,6].
Several studies have identified patient populations at an increased risk for postoperative residual ND. Reported clinical risk factors include preoperative neurological status, advanced age, diabetes, methicillin-resistant Staphylococcus aureus infection, and time to surgery [7–9]. Additionally, certain magnetic resonance imaging (MRI) features, such as abscess length and thickness, canal compromise ratio, and contrast enhancement patterns, have been correlated with poorer postoperative neurological outcomes [10–12]. A study confirmed these findings by identifying radiological risk factors, including abscess length, thickness, and canal compromise ratio, as predictors of residual ND [13]. However, these indicators have not yet been established as optimal predictors.
The Spinal Infection Treatment Evaluation (SITE) score was recently introduced as a comprehensive tool to evaluate five critical domains for predicting postoperative improvement in patients with de novo spinal infections requiring surgery [14–16]. One study demonstrated that no patient experienced postoperative neurological deterioration. However, the correlation between the SITE score and postoperative ND remains unexplored [15]. We believe that the SITE score is associated with postoperative neurological outcomes; however, no study has examined this association.
Therefore, this study evaluated the validity of the SITE score for predicting residual ND after surgery for SEA. By analyzing patients’ clinical characteristics, preoperative and postoperative neurological symptoms, MRI findings related to abscess involvement, and individual components of the SITE score, we aimed to refine preoperative assessments and enhance clinical decision-making.
Materials and Methods
Patient selection
We retrospectively reviewed the medical records of patients diagnosed with de novo SEA in the thoracic or lumbar spine who underwent posterior-only surgery at Taipei Veterans General Hospital between January 2005 and December 2014. Spinal infection was suspected in patients presenting with low back pain, fever of unknown origin, and/or newly developed ND, accompanied by elevated C-reactive protein and/or erythrocyte sedimentation rate. MRI was then performed to evaluate SEA, with gadolinium-enhanced MRI used for diagnosis. Preoperative computed tomography-guided biopsies were conducted on samples from the infected disc space, endplate, or vertebral body for culture and histopathology, except in cases that required emergency surgery. The inclusion criteria were as follows: MRI-confirmed de novo SEA involving the thoracic (T1–T12) or lumbar (L1–L5) spine, surgery via a posterior-only approach, and a minimum follow-up duration of 2 years. Patients with previous spinal surgeries, spinal infections, or incomplete medical records were excluded from the study. This study was approved by the Institutional Review Board (IRB) of the Taipei Veterans General Hospital (IRB number: 2025-01-014AC). All participants provided written informed consent in accordance with the Declaration of Helsinki.
Surgical indications
Surgical indications included ND, mechanical instability, tissue culture for diagnosis, and failure of medical treatment. Surgery was performed immediately in patients with ND. Urgent surgery, defined as surgery conducted within 24 hours of admission, was performed in patients without deficits to prevent potential neurological deterioration and optimize clinical outcomes. For patients with mechanical instability, posterior decompression with long posterior instrumentation (extending two levels above and below the lesion) and short posterolateral fusion (one level above and below the lesion) were performed. In the absence of instability, only posterior decompression was performed [17]. All procedures employed a posterior-only surgical approach, which has proven effective in managing thoracic and lumbar SEAs [18]. A single surgeon performed all surgeries, minimizing potential variability in surgical technique and experience.
Antibiotic treatment protocol
Based on pathogen identification, cases were classified as pyogenic, tuberculosis (TB), or hematogenous infections. For pyogenic infections, intravenous antibiotics were administered postoperatively for 6 weeks or until the C-reactive protein and erythrocyte sedimentation rates normalized, followed by 6 weeks of oral antibiotics. This regimen was overseen by the hospital’s infectious disease specialists to ensure precise management. Regarding TB, once the diagnosis was confirmed, rifampicin, isoniazid, ethambutol, and pyrazinamide were administered for 2 months, followed by a 10-month regimen of rifampicin, isoniazid, and ethambutol. Pulmonology specialists at our hospital supervised the TB treatment protocol in accordance with national guidelines [19].
Clinical and MRI risk factors
We collected data on demographic characteristics, including age, sex, body mass index (BMI), and comorbidities, such as diabetes, chronic renal disease, liver disease, and malignancy. Clinical information, including smoking habits and cauda equina syndrome, was also documented. We recorded the interval between symptom onset and surgery, which was defined as the time from the initial onset of back pain to the date of surgery. The SITE score, which integrates various factors such as neurology, location, radiology, pain, and host comorbidities, was calculated for each patient. Supplement 1 summarizes the components of the SITE score.
We analyzed radiological factors based on MRI findings, including the affected spinal level (thoracic, thoracolumbar junction, lumbar, and lumbosacral), number of involved vertebral levels, presence of an anteroposterior (AP) epidural abscess, canal compromise AP ratio, canal compromise cross-sectional area (CSA) ratio, abscess length, and abscess thickness. Radiological measurements were performed according to previously established protocols [13]. Neurological status was evaluated using the Frankel grading system before surgery, as well as at 2 and 6 weeks and 3 and 6 months after surgery; thereafter, it was assessed annually [20]. Patients were categorized into two groups based on their postoperative neurological status: those with residual ND (ND group) and those without residual ND (NND group).
We compared the clinical and radiological risk factors between the ND and NND groups. Logistic regression analysis was used to identify significant predictors of residual ND, including the overall SITE score. Additionally, a regression analysis of the SITE score components (neurology, location, radiology, pain, and host comorbidities) was conducted to identify the specific components that significantly predicted residual ND.
Statistical analysis
Statistical analyses were conducted using IBM SPSS software ver. 22.0 (IBM Corp., Armonk, NY, USA). Chi-square or Fisher’s exact tests were used for categorical variables, while continuous variables were analyzed using the Mann-Whitney U test. The significance threshold was set at p<0.05. Univariate and multivariate logistic regression models were used to identify factors associated with residual ND, with results expressed as odds ratios (ORs) and 95% confidence intervals (CIs). A cutoff SITE score of ≤8 was used as a threshold value indicating the need for surgery, as reported in a previous study [14]. Other risk factors, including age ≥70 years, canal compromise AP ratio ≥50%, canal compromise CSA ratio ≥57%, abscess length ≥5.5 cm, and abscess thickness ≥0.8 cm, were adopted from a study that established these thresholds using receiver operating characteristic (ROC) curve analysis [13]. Variables with p<0.1 in the univariate analysis were selected for inclusion in the multivariate model. The ROC curve and Youden index were employed to determine the optimal cutoff SITE score, and the sensitivity, specificity, and area under the curve (AUC) were calculated. A power analysis was conducted to ensure the adequacy of the statistical tests, with an α level of 0.05.
Results
The study included 45 patients diagnosed with SEA, with an average age of 67.2±13.1 years and a mean BMI of 24.4±4.7 kg/m2. The cohort comprised 69% male (n=31) and 31% female (n=14) patients. Smoking habits were reported in 15.6% of patients (n=7). Comorbidities, including diabetes (28.9%), chronic renal disease (8.9%), liver disease (17.8%), and malignancy (11%), were prevalent. The thoracic (20%), thoracolumbar junction (8.9%), lumbar (51.1%), and lumbosacral (20%) levels were affected. The mean number of involved vertebral levels was 2.8±1.0. The mean interval between symptom onset and surgery was 16.6±20.2 weeks. Cauda equina syndrome was present in 24.4% of the patients. The average SITE score was 6.1±2.1 points, and the mean follow-up period was 41.4±26.9 months (Table 1).
Patient characteristics
Neurological status, assessed using the Frankel grading system, showed significant improvement from the preoperative period to the final follow-up examination. Preoperatively, the neurological status of the patients was distributed across Frankel grades B to E, with 9%, 22%, 31%, and 38% classified as grades B, C, D, and E, respectively. No patients were classified as Frankel grade A. At the final follow-up, the distribution shifted toward improved functional outcomes, with 2%, 4%, 27%, and 67% of the patients classified as grades B, C, D, and E, respectively. No patient was classified as Frankel grade A at any time point, highlighting the substantial postoperative recovery of neurological function (Fig. 1).
Comparison of the postoperative neurological status with that at the final follow-up for patients who underwent surgery for spinal epidural abscess. Values are presented as number (%).
The clinical characteristics significantly differed between the ND (n=15) and NND (n=30) groups. The SITE score was significantly lower in the ND group (4.3±1.3) than in the NND group (7.0±1.8; mean difference, 2.7; 95% CI, 1.6–3.7; p<0.0001). The ND group also exhibited a higher incidence of cauda equina syndrome (53% vs. 10%, p=0.0016) and a shorter interval from symptom onset to surgery (9.9±13.4 weeks vs. 20.0±22.0 weeks, p=0.019). Other clinical variables, including age, BMI, and comorbidities, such as diabetes and chronic renal disease, did not show significant between-group differences (Table 2).
Clinical risk factors of patients in the neurological deficit and non-neurological deficit groups
Logistic regression analysis was conducted to identify the risk factors associated with postoperative residual ND. In univariate analysis, significant predictors included the SITE score (OR, 2.39; 95% CI, 1.48–3.93; p<0.0001), A+P epidural abscess (OR, 6.0; 95% CI, 1.24–29.07; p=0.020), canal compromise AP ratio (OR, 12.25; 95% CI, 1.42–105.36; p=0.0036), abscess length ≥5.5 cm (OR, 5.50; 95% CI, 1.39–21.72; p=0.010), and abscess thickness ≥0.8 cm (OR, 17.88; 95% CI, 3.28–97.30; p<0.0001). In multivariate analysis, the SITE score remained the only independent predictor (OR, 2.70; 95% CI, 1.25–5.88; p=0.012) (Table 3).
Univariate and multivariate logistic regression of risk factors for postoperative residual neurologic deficits
Regression analysis of the SITE score components revealed that no individual component was a significant predictor in the multivariate analysis. However, in the univariate analysis, neurological (OR, 20.0; p=0.0001) and pain (OR, 14.29; p<0.0001) components were associated with residual ND. Despite this, neither component emerged as a significant predictor in the multivariate analysis (Table 4). To determine whether multicollinearity or model misspecification affected the significance of the SITE score, we calculated the variance inflation factor (VIF). Both VIF values were <5 (neurology: 1.73; pain: 1.73), indicating no significant multicollinearity.
Logistic regression analysis of individual SITE score components for predicting residual ND (n=45)
ROC curve analysis demonstrated that a SITE score of ≤6 was highly predictive of residual ND, with a sensitivity of 100.0% (95% CI, 78.2–100.0) and a specificity of 73.3% (95% CI, 54.1–87.7; AUC, 0.877; 95% CI, 0.744–0.956; p<0.001). Patients with a SITE score ≤6 had an OR of 82 for developing residual ND compared to those with a SITE score >6. These findings underscore the reliability of the SITE score in predicting postoperative neurological outcomes (Fig. 2). The distribution of patients stratified by SITE scores ≤6 and >6 is presented in Table 5.
Receiver operating characteristic curve for the Spinal Infection Treatment Evaluation (SITE) score as a predictor of postoperative neurological deficits. The Youden index is used to determine the optimal cutoff value, yielding a sensitivity of 100.0% and a specificity of 73.3%. A SITE score ≤6 indicates a significantly higher risk (odds ratio [OR], 82) of postoperative neurological deficits. AUC, area under the curve.
Distribution of patients stratified by SITE score ≤6 and >6
Discussion
To our knowledge, this is the first study to validate the value of the SITE score in assessing neurological outcomes after surgery for SEA. Patients with a SITE score ≤6 were more likely to experience postoperative ND than those with a SITE score >6. Previous studies identified specific clinical and radiological indicators associated with postoperative ND [7,8,10–13]; however, this study involved a comprehensive assessment of the overall SITE score as a predictor that may complement other clinical and radiological indicators.
This study utilized a cohort previously reported in a separate analysis of risk factors [13], with stricter inclusion criteria for de novo spinal infection, aligning with the original assumptions of the SITE score. However, in the subsequent multivariate analysis of risk factors (Table 3), significant predictors of residual ND reported in a previous study [13], such as age and abscess thickness, lost significance. This change may be attributed to the introduction of a more critical variable (the SITE score), which effectively accounted for the variability in the outcome measures. The relevance of other factors was diminished compared to that of the SITE score.
Given that conservative management is recommended for patients with a SITE score ≥8, we compared patients with a SITE score >8 to those with a SITE score <8 and found no significant difference in neurological outcomes between these two groups. This suggests that a score of 8 points cannot simultaneously determine the necessity of surgery and the likelihood of residual ND. We found that a SITE score of 6 was the most appropriate cutoff value, as patients scoring ≤6 points had an 82-fold higher risk of postoperative ND than those scoring >6 points.
Moreover, although the SITE score includes an ND component, our analyses of individual components in isolation revealed that none were significantly correlated with neurological outcomes. These results suggest that predicting neurological outcomes requires a comprehensive assessment of the patient’s condition rather than focusing on a single aspect. Therefore, integrating multiple components into the SITE score is crucial for accurately predicting postoperative ND.
Two studies by Pluemer et al. [14,16] focused on calculating the SITE score and analyzing its predictive performance compared to the Spinal Instability Spondylodiscitis Score (SISS) and the Spinal Instability Neoplastic Score (SINS). Using an expert panel decision approach, these studies demonstrated that the SITE score had higher sensitivity and specificity than SISS and SINS. However, they did not examine the relationship between the SITE score and surgical outcomes. Similarly, Rezvani et al. [15] reported significant postoperative improvements in the Visual Analog Scale scores for back pain and subscale scores for the Japanese Orthopedic Association Back Pain Evaluation Questionnaire among patients undergoing surgery, with no cases of postoperative neurological deterioration recorded using the American Spinal Injury Association grading. However, that study did not analyze the association between the SITE score and postoperative outcomes, including residual ND.
Previous studies have suggested that a shorter interval between abscess diagnosis and surgery is associated with a lower likelihood of residual ND [9]. Other studies have indicated that surgical intervention should be implemented promptly to avoid prolonged hospitalization and consequent in-hospital mortality [21,22].
In contrast, a longer duration from symptom onset to surgery was associated with fewer neurological sequelae in our patient group. This may be attributed to our definition of symptom onset as the initial onset of back pain. Patients who underwent delayed surgery tended to have milder preoperative symptoms, which likely allowed for a longer interval before surgery.
Despite the significant implications of our findings, this study had some limitations. First, this was a retrospective analysis with a relatively small sample size, which may limit the generalizability of the results. In addition, the power analysis results fell below the predefined threshold of 0.8. Second, the study was conducted at a single center; future research should include multicenter data to further validate the reliability of the findings. Third, the lack of significance for the individual components of the SITE score in the multivariate analysis suggests that large-scale studies are warranted to confirm the predictive value of these components. Finally, variations in surgical procedures based on the patient’s condition, type of pathogen, and operating surgeon may have introduced observational bias.
Future research should focus on increasing the sample size and conducting prospective studies to further verify the predictive utility of the SITE score. Additionally, studies should explore ways to integrate the SITE score into existing clinical pathways and assess its applicability across diverse healthcare settings. Future studies should also consider evaluating additional predictive markers to enhance existing models.
Conclusions
This study demonstrated that the SITE score is a reliable and independent predictor of residual ND after surgery for SEA. Patients with lower SITE scores, particularly those with a score ≤6, may have a significantly higher risk of postoperative ND. These findings underscore the importance of the SITE score in preoperative assessment and suggest its potential utility in guiding clinical decisions and improving patient outcomes.
Key Points
The Spinal Infection Treatment Evaluation (SITE) score effectively guides surgical decisions in cases involving de novo spinal infections.
The SITE score reliably predicts neurological deficits after surgery for spinal epidural abscess (SEA).
SITE scores ≤6 indicate a higher risk of postoperative neurological deficits.
The individual components of the score lack predictive value; therefore, a holistic approach is essential.
Notes
Conflict of Interest
No potential conflict of interest relevant to this article was reported.
Acknowledgments
We thank and acknowledge the contributions of all participants and team members. The data used in this article will be shared upon reasonable request from the corresponding authors.
Author Contributions
Conception and design of the study: JJC, HHL, PHC. Data acquisition: JJC, HHL, PHC. Data analysis: JJC, HHL, PHC. Interpretation: JJC, HHL, PHC, STW, CLL. Manuscript drafting: YCY. Critical revision of the manuscript: STW, CLL. Supervision: YCY. Final approval of the manuscript: all authors
Supplementary Materials
Supplementary materials can be available from https://doi.org/10.31616/asj.2025.0285.
Supplement 1. The SITE score
asj-2025-0285-Supplementary-1.pdf