Spinal Epidural Abscess: Early Suspicion in Emergency Department Using C-Reactive Protein and Erythrocyte Sedimentation Rate Tests

Abstract

Background

Spinal epidural abscess (SEA) is an uncommon but serious differential diagnosis of acute spinal pain with high paralysis and mortality rate. This study aims to provide local data on its bioclinical characteristics and evaluate potential strategies to enhance its diagnostic rate in accident and emergency department (AED).

Methods

A retrospective case study from 2013 to 2019 was conducted in United Christian Hospital. SEA cases were classified as study group, spinal pain due to non-SEA cases were classified as control group. Data collected from study group included symptoms, radiological diagnosis, microbiological culture, treatment, and outcome. C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) of both groups were compared to analyze their diagnostic power.

Results

In the study group (n = 42), 93% of patients had spinal pain, 55% had fever, 60% had neurological deficits, and only 26% had the classic triad on presentation. Seventy-four percent of patients presented with spinal cord or cauda equina compression in their first magnetic resonance imaging. Mortality rate was 23.8%, and paralysis rate was 7.1%. Diagnostic accuracy in AED was 12%. Admission to orthopedic ward (n = 23) resulted in a significantly lower mean time-to-imaging (4.39 days vs. 14.58 days) and mean time-to-treatment (6.56 days vs. 16.9 days) as compared to other specialties. The area under curves of CRP and ESR were 0.893 and 0.874 respectively, the optimal threshold levels were 45.9 mg/L (sensitivity 82.9%, specificity 79%) and 59.5 mm/hr (sensitivity 87.2%, specificity 80.4%), respectively.

Conclusion

Diagnosis of SEA in emergency department based on clinical symptomatology is not reliable due to low incidence of the classic triad, despite a more advanced disease on presentation. We proposed incorporating CRP and ESR tests into evaluation of patients with spinal pain since both tests demonstrated excellent discriminative power in diagnosing SEA.

Introduction

Low back pain and neck pain are common symptoms in patients presenting to accident and emergency department (AED). Among their wide spectrum of differential diagnoses, spinal epidural abscess (SEA) is an uncommon but potentially devastating and fatal condition. Its incidence is low, around 5.1 cases per 10,000 hospital admission, 1 despite being much higher compared to 0.2–2.0 cases per 10,000 hospital admission in the 20th century 2 due to aging population and improving availability of magnetic resonance imaging (MRI). However, its morbidity and mortality rate remain high, causing around 8%–15% paralysis and 7%–16% death. 1 , 3

One of the reasons accounting for the poor outcome is due to its notoriously ambiguous symptomatology. It classically presents with a triad of spinal pain, fever, and neurological deficits, but studies have shown that the triad is only present in around one-third of patients. 4 As a result, initial misdiagnosis based on clinical signs was common on presentation to AED, and patients were often labelled to have intervertebral disc prolapse, spinal spondylosis, etc. 3 Without timely administration of systemic antibiotics and surgical decompression, the disease often progresses rapidly from stage 1 (back pain, fever) and stage 2 (spinal radicular pain, hyper-reflexia) to more advanced stage 3 (sensory, motor and sphincter deficit) and stage 4 (paralysis). 5 This caused serious morbidity since preoperative motor deficit and patient age were found to be significantly associated with residual neurological outcome. 6 Therefore, there is a dire need to improve diagnostic accuracy of SEA in AED to facilitate early treatment in order to improve its outcome. However, studies addressing this issue remain sparse.

This study aims to provide data on bioclinical characteristics of local SEA cases in Hong Kong, investigate how a decision in AED would impact the time to diagnosis and treatment as well as clinical outcome, and evaluate its diagnostic accuracy by C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) in AED.

Methods

We performed a retrospective study of SEA in adults from January 1, 2013 to December 31, 2019 at United Christian Hospital in Hong Kong. Potential cases were identified through searching the Clinical Data Analysis and Reporting System of Hospital Authority by International Classification of Diseases, Ninth Revision (ICD-9) codes and their data extracted from Electronic Patient Record.

Adult patients (older than 18 years) with SEA (ICD-9 324.1, 13.5) were classified as study group. The following key information was documented:

1. Patient demographics: gender, age, and hospital number.

2. Symptoms: spinal pain, fever (defined as temperature ≥ 38°C), and neurological deficits (defined as any sensory disturbance, reduced limb power, sphincter impairment).

3. Decision made in AED: initial diagnosis, admission, or discharge destination.

4. Investigations: CRP, ESR, MRI diagnosis, blood or pus culture result.

5. Treatment: surgical operation or image-guided drainage.

6. Outcome: time to MRI diagnosis, time to definitive treatment, length of stay in acute ward and ambulatory status on discharge.

Adult patients with spinal pain caused by non-SEA diagnoses were classified as control group. The cases were captured by diagnostic codes, such as backache, sciatica (ICD-9 724), spondylosis, spondylolisthesis (ICD-9 721), intervertebral disc disorder (ICD-9 722), inflammatory spondylopathies (ICD-9 720), malignant neoplasm of vertebral column (ICD-9 170.2), etc.

Continuous data were presented as mean and standard deviation (SD). A univariate analysis was performed by independent Student’s t -test for continuous variables. Statistical analysis was calculated by IBM Statistical Product and Service Solutions (SPSS ^® ) with its corresponding p value and 95% confidence interval (CI) illustrated. By comparing the CRP and ESR levels of both study and control group, their diagnostic sensitivity and specificity were calculated. Receiver operator characteristic (ROC) curves were generated with SPSS ^® , and the optimal threshold point was determined by Youden Index (sensitivity + specificity –1). MRI was used as the reference standard.

Results

In the study group, a total of 42 SEA cases were identified during the 7-year study period from 2013 to 2019. The male-female ratio was 7:3. The mean age was 67 years (range: 27–91, SD: 15.59).

On presentation to AED, 93% of patients reported spinal pain, 55% reported fever, and 60% reported neurological deficits. The classic triad was only present in 26% of cases.

MRI after admission demonstrated that 40% of SEA involved lumbar level, 26% involved cervical level, 21% involved thoracic level, and 12% involved thoracolumbar level. Notably, radiological spinal cord or cauda equina compression was present in 74% of cases in their first MRI.

A causative pathogen was identified in culture of aspirated pus or blood in 79% of patients. Staphylococcus was the most common (31%, 13 patients), in which 12 were methicillin-sensitive, and 1 was methicillin resistant. Streptococcus was the second most common (19%), while Mycobacterium tuberculosis was the third most common (14%).

Twenty-five patients (60%) were managed with surgical operation, in which 14 patients received decompressive laminectomy only, whereas the other 11 patients received additional spinal fusion. Seventeen patients (40%) were managed conservatively with medical treatment, in which 14 patients received systemic antibiotics only, whereas the other 3 patients received additional computed tomography (CT) guided drainage. One patient received both surgical operation and CT-guided drainage ( Table 1 ).

Table 1 . Characteristics of study group .

CT: computed tomography; MRI: magnetic resonance imaging; SD: standard deviation.

Characteristic	Patient no. (%)
Patient demographics	42
Mean age, years (SD)	67 (15.59)
Male	32 (76)
Female	10 (24)
Symptoms
Spinal pain	39 (93)
Fever	23 (55)
Neurological deficits	25 (60)
Classic triad	11 (26)
MRI level
Cervical	11 (26)
Thoracic	9 (21)
Thoracolumbar	5 (12)
Lumbar	17 (40)
Causative organism
Staphylococcus aureus	13 (31)
Streptococcus	8 (19)
Mycobacterium tuberculosis	6 (14)
Enterococcus	2 (5)
Escherichia coli	2 (5)
Staphylococcus epidermidis	1 (2)
Klebsiella	1 (2)
Negative culture	9 (21)
Treatment
Surgical treatment	25 (60)
Laminectomy only	14
Laminectomy with spinal fusion	11
Medical treatment	17 (40)
Systemic antibiotics only	14
CT guided drainage	4

The mean length of stay in acute ward was 55.2 days (range: 10–221, SD: 37.52). On discharge, 69% of patients were ambulatory, 7.1% were paralyzed. Overall mortality rate was 23.8%. Notably, 63.6% of patients older than 80 years old succumbed, while all patients younger than 60 years old survived ( Table 2 ).

Table 2 . Age distribution and outcome of study group ^a .

^a Data are reported as number (%).

Age (years)	Ambulatory	Paralysis	Mortality
18–59	11 (92)	1 (8)	0 (0)
60–79	15 (79)	1 (5)	3 (16)
≥ 80	3 (27)	1 (9)	7 (64)

There were 23 patients (55%) admitted to orthopedics ward. Considering phrases like “spinal epidural abscess, spinal infection, spondylodiscitis” as relevant diagnoses, the diagnostic accuracy of SEA in AED was only 12%. The mean time from admission to radiological diagnosis by MRI was 9 days (range: 0–41, SD: 10.02). Early clinical diagnosis of SEA in AED caused a significantly lower time-to-MRI (mean difference: 8.173 days, 95% CI: 3.64–12.71, p = 0.001). Orthopedic admission of SEA patients in AED caused a significantly lower time-to-MRI (mean difference: 10.19 days, 95% CI: 4.05 to 16.32, p = 0.002) and time-to-definitive treatment (mean difference: 10.34 days, 95% CI: –0.48 to 21.17, p = 0.059) ( Table 3 ).

Table 3 . Mean time and outcome comparison between admission to orthopedic and non-orthopedic ward .

CI: confidence interval; MRI: magnetic resonance imaging.

Characteristic	Orthopedic admission	Non-orthopedic admission	Mean difference (95% CI)	p value
No. of patients (%)	23 (55)	19 (45)
Time
to MRI (days)	4.39	14.58	10.19 (4.05 to 16.32)	0.002
to treatment (days)	6.56	16.90	10.34 (–0.48 to 21.17)	0.059
length of stay (days)	57.78	52.11	5.67 (–17.36 to 28.71)	0.621

In control group, a total of 2,223 patients were identified with CRP test done (42% male, 58% female), 4,272 patients with ESR test done (38% male, 61% female). The respective mean age was 70.3 and 70.9 years old.

The mean CRP in study group was 149.9 mg/L (range: 11.7–432.7 mg/L), which is 4.0 times higher compared with 30.1 mg/L in control group (mean difference: 119.8, 95% CI: 89.1–150.6, p < 0.001) ( Table 4 ). The area under curve of CRP test is 0.893 (95% CI: 0.850–0.935). The optimal threshold level is 45.9 mg/L, giving an 82.9% sensitivity and 79.0% specificity ( Fig. 1 ).

Table 4 . Comparison of mean CRP and ESR levels .

CI: confidence interval; CRP: C-reactive protein; ESR: erythrocyte sedimentation rate.

Characteristic	Study group	Control group	Mean difference (95% CI)	p value
Sample size	n = 41	n = 2,223
Mean CRP (mg/L)	149.9	30.1	119.8 (89.1–150.6)	< 0.001
Sample size	n = 39	n = 4,272
Mean ESR (mm/hr)	86	37	49 (39.6–58.1)	< 0.001

Fig. 1 . Receiver operating characteristic (ROC) curves of C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) test. ROC curves for predicting spinal epidural abscess using CRP (blue continuous line) and ESR (red dotted line) tests were plotted and their area under curve was shown in the box. The dots on the curve represented the optimal threshold value of both tests, and their corresponding sensitivity (sen) and specificity (spec) were shown.

[The colored figure is used in the online version, available on the website http://doi.org/10.6705/j.jacme.202303_13(1).0003]

The mean of ESR in study group was 86 mm/hr (range: 7–127 mm/hr), which is 1.32 times higher compared with 37 mm/hr in control group (mean difference: 49, 95% CI: 39.6–58.1, p < 0.001) ( Table 4 ). The area under curve of ESR test is 0.874 (95% CI: 0.817–0.931). The optimal threshold level is 59.5 mm/hr, giving an 87.2% sensitivity and 80.4% specificity ( Fig. 1 ).

Discussion

Bioclinical Characteristics of Local Cases

This study reported a case series of 42 patients with SEA in a local hospital, and the majority were elderly male. Almost all patients reported spinal pain (93%), more than half reported fever (55%), and neurological deficits (60%), yet the classic triad only occurred in one-fourth (26%) of patients. This resonated with the data from two largest case series to date including a total of 290 cases, which reported that 92%–100% SEA patients presented with spinal pain, 50.0%–62.4% with fever, 42%–47% with neurological deficits, and 33% with the classic triad. 7 The low frequency of the classic triad demonstrated the difficulty in diagnosing SEA clinically based on symptoms and signs.

Although the majority (74%) of SEA occurred in thoracolumbar level presenting as back pain, there was also a considerable portion (26%) occurring in cervical level presenting as neck pain and nuchal rigidity with low grade fever, resulting in it often being mistaken as meningitis by emergency physicians. Its lower incidence and non-specific presentation were similarly reported by the only systemic review in recent years focusing specifically on cervical SEA, which in addition showed that it resulted in higher risk of morbidity and mortality. 8

The ratio of surgical to medical treatment was six to four. Common reasons for taking a conservative treatment approach were due to poor premorbid state, patient declining surgery, absence of compressive neurological deficit, lesion too small for drainage, or lesion at atlantoaxial level. A similar ratio was reported in a systemic review comparing medical and surgical management of SEA, which commented that there has been a substantial trend toward treating neurological intact patients with medical treatment. 9 Multiple studies have emerged recently to evaluate predictors of failure of nonoperative management based on risk factors such as the presence of diabetes mellitus, and laboratory data such as leukocytes and CRP level, 10 potentially guiding prioritization and redistribution of treatment resources in future cases.

Mortality rate of SEA in this study was 23.8%, which is higher compared to 16% as reported by the largest meta-analysis to date. 3 However, 70% of the deceased were elderlies older than 80 years old who were mostly frail with multiple comorbidities, and they eventually died of complications such as pneumonia and acute myocardial infarction. This was supported by a machine learning algorithm for prediction of mortality of SEA proposed in one recent study, which concluded that age was the most important factor as compared to other factors such as albumin, platelet count, active malignancy, and diabetes. 11 In contrast, all patients younger than 60 years old survived in our study, and 92% of them were ambulatory on discharge.

Importance of Early Suspicion

Its ambiguous symptomatology reflects the difficulty to make a clinical diagnosis of SEA based on symptoms and signs in AED. It may be one of the reasons accounting for the low diagnostic accuracy (12%) and orthopedic admission (55%) of SEA cases in our study. The most common scenario was admitting a non-communicable old age home resident to medical ward due to fever without localizing sign or fall because of lower limb weakness. This caused a delay in mean time to MRI diagnosis to more than a week (mean 9 days). Similar situation had been observed in the emergency department of UC San Diego Medical Centre, which reported a case series of 63 patients with diagnostic delays present in 75% of patients, and the median number of ED visits before SEA diagnosis was two. 12 In a 10-year study conducted at Baystate Medical Centre that included 162 cases, more than half of the cases have had at least one episode of medical visit before definitive diagnosis and admission. 13

Devastating consequences may be due to a subsequent delay in definitive treatment and surgery since irreversible paralysis commonly occurs 48 hours after cord or cauda equina compression, which was present in the majority of patients (74%) in our case series. A study conducted in Massachusetts General Hospital of Boston found that SEA patients with high grade cord compression (based on its proposed novel 5-point grading system) were significantly more likely to present with pretreatment neurological deficit. 14 Although the study did not find any association between the severity of cord compression with residual motor deficit and 90-day mortality, multiple studies have shown that preoperative motor deficit was significantly associated with poorer neurological outcome, 3 , 6 , 15 and diagnostic delays in emergency department had a higher proportion of residual motor weakness (45%) than those without (13%). 12

Recommendation on Improving Diagnostic Accuracy

Therefore, it is utmost important to achieve early clinical diagnosis of SEA, despite its difficulty, as this study demonstrated that orthopedic admission or early clinical diagnosis of SEA cases in AED significantly reduce the mean time to MRI diagnosis and definitive treatment.

In this study, we found that the mean value of CRP and ESR test were significantly higher when comparing the study group of SEA patients with the control group of patients suffering from non-SEA causes of back pain. Both tests showed excellent discriminative power in diagnosing SEA as their area under curves were larger than 0.8. Their optimal threshold points demonstrated excellent sensitivity and specificity (approximately 80%) at 45.9 mg/L for CRP test and 59.5 mm/hr for ESR test.

Hence, we proposed the following diagnostic algorithm of SEA that incorporated the use of CRP and ESR locally in Hong Kong ( Fig. 2 ).

Fig. 2 . Diagnostic algorithm of spinal pain in accident and emergency department. This flowchart illustrated a step-by-step diagnostic evaluation of spinal pain patients. First, presence of spinal pain with alarming neurological deficits or fever justifies direct admission to orthopedics ward. If both symptoms are absent, perform C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) test in patients who have risk factors for spinal epidural abscess (SEA), suspicious spinal X-ray findings or are unresponsive to analgesics. Admit if CRP or ESR are above their optimal threshold values.

MRI: magnetic resonance imaging.

Similar approach of incorporating CRP and ESR test into a diagnostic algorithm based on risk factors and clinical symptoms were used in multiple studies. A recent systemic review written by Tetsuka et al. 16 in 2020 proposed a step-by-step algorithm. First, examine any neurological deficit. Second, evaluate any risk factors or presence of fever. Third, perform CRP and ESR test. If any diagnostic step give a positive result, then arrange urgent MRI immediately without proceeding to next step. Similar diagnostic algorithms were used in other studies, for example, Bond and Manian 17 emphasized evaluating any recent staphylococcus bacteremia or spinal instrumentation in step two, Alerhand et al. 18 focused on performing ESR only on step three. One similar algorithm was put into test in a cohort analysis, comparing the incidence of diagnostic delays and presence of motor deficits before and after implementing it. The study showed a significant reduction in both diagnostic delays (83.6% before vs. 9.7% after) and motor deficits (81.8% before vs. 19.4% after). 19

Limitations

Limitation of this study included small sample size, short review period, and limited experience in a single local hospital. We were unable to demonstrate that orthopedic admission or early diagnosis of SEA produced any significant difference in hospital length of stay, paralysis, or mortality rate. One of the possible explanations is that patients who presented to AED with prominent symptoms severe enough to be noticed by emergency physician often have more advanced disease courses, whereas patients with mild non-specific symptoms that were misdiagnosed in AED may have their disease in early stage. Also, when evaluating the specificity of CRP and ESR, we did not include non-musculoskeletal differential diagnoses of spinal pain such as acute aortic syndromes or other infective causes (e.g., pyelonephritis, biliary sepsis) in control group as they were beyond the scope of our study. We believe that these diagnoses, which certainly would cause an elevated CRP and ESR result, should be considered and ruled out clinically before adopting any tests or algorithms suggested in this study. In other words, the high specificity of CRP and ESR test in diagnosing SEA only holds true in patients with musculoskeletal spinal pain.

Conclusion

In conclusion, this is a retrospective cohort study of 42 SEA cases in 7 years in a local hospital. Its high mortality rate, especially in the elderlies, signified the importance of its recognition among patients presented with acute spinal pain. This study demonstrated that early diagnosis and orthopedic admission of SEA cases in AED significantly reduce the mean time to MRI diagnosis and definitive treatment. However, the ambiguity of its symptomatology caused a low diagnostic accuracy in AED. Therefore, we proposed the addition of CRP and ESR as screening tests in evaluating patients with spinal pain. Unique to previous studies in current literature, this study was able to produce optimal threshold values for both tests based on analysis of their ROC curves, and we concluded that they demonstrated excellent discriminative power with great sensitivity and specificity in identifying SEA cases. We suggest future research to compare the outcome difference of SEA before and after incorporating CRP and ESR tests into a new diagnostic algorithm when evaluating patients with spinal pain in AED.

Conflicts of Interest Statement

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Acknowledgments

I would like to thank my seniors Dr. Chan Yiu Cheung, Dr. Raymond Wong for their guidance and support. I would also like to thank my department executive assistant Mr. Tse Kar Wai for his support on statistical calculations.