Introduction

Spinal epidural abscesses (SEAs) represent rare but serious infections with significant repercussions. Reported mortality rates reach up to 20%, and many survivors experience severe neurological sequelae that impair their quality of life [1,2].The neurological deficits resulting from SEAs can become irreversible within 48 hours of onset [3,4]. Therefore, timely diagnosis and effective treatment are essential to prevent possible complications and achieve adequate resolution, minimizing possible sequelae.

Nonetheless, the optimal therapeutic strategy (only antibiotics vs a combination of antibiotics plus drainage intervention) for these patients is not well defined. There is still controversy about the need for surgery in all patients with SEA [5–7].Historically, SEA was regarded as an absolute indication for drainage intervention [8,9]. However, over the last two decades, small case series and cohorts have reported favourable outcomes in certain patients treated medically (antibiotics only) [10–13]. Consequently, many experts now recommend an initial medical approach in patients without neurological deficits [14,15].

Nevertheless, the current evidence supporting the efficacy and safety of an antibiotic-based treatment for SEA is controversial. Other cohorts have reported a high rate of therapeutic failure with impact in both quality of life and neurological motor sequelae [5,16–21]. Reflecting this uncertainty, a recent Delphi consensus recommendations failed to reach agreement on whether SEA’s represent a surgical indication regardless the presence of neurological deficits [22].

In this context, we aimed to describe the rate of clinical failure after initial medical management of patients with SEA and to compare the outcomes of these patients to those who underwent early drainage intervention. We also aimed to identify factors associated with failure after medical treatment.

Materials and methods

We conducted a single-center retrospective cohort at a tertiary hospital. Our center is a 700-bed university hospital serving a reference population of over 500,000 inhabitants. All types of complex spinal surgery are performed.

We included all adult patients (over 18 years) diagnosed with SEA between January 2009 and December 2022. Patients with isolated SEA and those with associated spondylodiscitis were eligible. There were no exclusion criteria.

Epidemiological, clinical, microbiological, radiological and prognostic variables were obtained from the electronic medical records using an anonymized form. Data management was performed using REDCap electronic capture tools, with licenses provided to the Research Institute Puerta de Hierro-Segovia de Arana [23,24]. The study was approved by the Hospital Puerta de Hierro Ethics Committee (code PI 240/25). Owing to its retrospective design, the requirement for informed consent was waived. The study was conducted in accordance with the recommendations of the Declaration of Helsinki.

Results

A total of 76 patients with SEA were included. Of these, 17 (22.4%) received initial intervention, whereas 59 (77.6%) received initial medical management. Among the initial interventions, 15 (88.2%) involved open surgery, and two (11.8%) involved drainage through interventional radiology.

Among medically managed patients, 25 (42.4%) met the criteria for failure. In comparison, treatment failure occurred in 3 patients (17.6%) within initial intervention group (p = 0.053); 1 due to SEA-related death and 2 due to neurological deterioration.

Spinal epidural abscess description

Table 1 summarizes the baseline demographic, clinical, microbiological and radiological characteristics of the cohort. The median age was 67 years (RIQ 57–76), 35.5% were women (27/76), and the median Charlson comorbidity index was 1 (RIQ 0–3). Twelve patients (15.8%) had prior spinal surgery, and seven (9.2%) had spinal instrumentation.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Univariate comparison of patients with initial intervention versus initial medical treatment.

https://doi.org/10.1371/journal.pone.0348373.t001

The median duration of symptoms until diagnosis was 2 weeks (IQR 1–4). Only 52.6% (40) had fever, and 17.1% (13) presented motor neurological impairment. Bacteremia was documented in 52.0% (39/75) of the patients. The most frequent etiologies were Staphylococcus aureus (46.1%) and coagulase-negative staphylococci (13.2%), followed by gram-negative bacteria (9.2%) and streptococci (6.6%). A polymicrobial etiology was found in 9.9% of the patients, whereas 21.1% (16) had no microbiological diagnosis.

Regarding radiological findings, SEAs were more common at the lumbar (47.4%) or thoracic levels (44.7%). Only 7.9% of patients had cervical involvement. In 53.4% of the patients, the lesions extended over more than one intervertebral space, contacted the spinal cord in 32% and produced radiological myelitis in 6.7%. A total of 13.2% of the patients had non-contiguous SEAs. In 69.4% of the patients, the abscess involved the anterior epidural area. In 90.2% (69/76) there was concomitant spondylodiscitis

The initial empirical antibiotic was appropriate in 93.2% of the patients. The median duration of antibiotic treatment was 57 days (IQR 45–99). In-hospital mortality was 10.5% (8), with SEA-related mortality accounting for 6 deaths (8.1%).

Factors associated with medical treatment failure

Among the patients initially managed medically (n = 59), 25 (42.4%) met criteria for failure. Sixteen patients (27.1% of the total, 64.0% of failures) underwent salvage surgery. The median time to treatment failure was 21 days (RIQ 14–40). The most frequent reasons for failure included radiological progression (n = 14), pain worsening (n = 13) and neurological deterioration (n = 7). The reasons for not performing an initial surgery in those with medical treatment failure were a surgical team’s decision according to neurological status and MRI findings in the vast majority (n = 24). Only 1 patient was considered not-suitable for surgery due to prohibitive comorbidity and poor baseline functional and cognitive status.

Of note, 8 patients presented with neurological motor impairment and were not initially operated. Of these, 2 patients had pre-existing motor deficit (1 of them failed medical treatment and were later operated), 2 had a deficit lasting more than 48 hours (1 of them also failed and were later operated), 1 had poor previous functional and cognitive status (died due to SEA) and 3 were classified as “mild” neurological deficit. These 3 patients failed medical treatment.

Table 2 compares the characteristics of patients who were successfully managed medically and those who failed medical treatment. Patients with failure more frequently presented diabetes mellitus (32.0% vs 8.8%, p = 0.040), a erythrocyte sedimentation rate (ESR) greater than 75 mm/h (66.7% vs 31.0%, p = 0.021), methicillin-resistant Staphylococcus aureus (MRSA) isolation (28.0% vs 0%, p = 0.001) and ventral epidural involvement (91.3% vs 60.6%, p = 0.014). There were no differences in other baseline characteristics, clinical presentation, microbiology or radiology between groups.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Univariate comparison of patients with initial medical management with and without treatment failure.

https://doi.org/10.1371/journal.pone.0348373.t002

Although rates of adequacy of empirical antibiotics were similar between groups, patients with medical treatment failure had longer intravenous antibiotic durations (26 days vs. 15, p = 0.033) and longer hospital admissions (46 days vs. 29, p = 0.012).

Primary and secondary endpoints

Table 3 shows the outcomes according to treatment group. The composite primary endpoint occurred significantly more frequently in patients who failed medical treatment (65.0%, 13/20) than in those without medical treatment failure (12.5%, 4/32, p < 0.001) or those treated initially with surgery (35.3%, 6/17, p = 0.035).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Comparison of prognosis and outcomes among patients with initial intervention, successful medical treatment and failure to medical treatment.

https://doi.org/10.1371/journal.pone.0348373.t003

Patients with medical treatment failure, compared with those without failure and those who initially underwent surgery, had higher SEA-related mortality (16.0% vs 0% vs 5.9%, p = 0.038). Neurological sequelae at discharge were also more common (31.3% vs. 6.3% vs. 25.0%, p = 0.013). Additionally, patients with medical treatment failure had worse ASIA scores at discharge (87 (SD 30) vs 99 (SD 2) vs 93 (SD 27), p = 0.035). They also showed greater deterioration in ASIAms from admission to follow-up (−3.8 (SD 6.5) vs + 0.6 (SD 2.1) vs + 4.8 (SD 8.5), p = 0.051).

Discussion

In our work, we aimed to describe the failure rate of medical management for SEAs and to evaluate the impact of such failure on clinical outcomes. Our main findings indicate that the failure rate of medical treatment is high (up to 40%) and could be associated with worse outcomes. In our cohort, factors associated with treatment failure included diabetes mellitus, an elevated ESR, MRSA isolation and ventral involvement of the abscess. Accordingly, initial surgery may be considered in those patients with these risk factors. In cases in whom the decision was made to treat only with antibiotic management, close monitoring during the first weeks is highly recommended.

In our series, three-quarters of the patients were initially managed without drainage intervention. This proportion is higher than that reported in both historical and recent series [25–27]. Nevertheless, this trend is consistent with the most recent recommendations from international societies, that propose a trial of medical management in neurologically intact patients [14,15]. A recent meta-analysis revealed that medical management is becoming increasingly common [6]. However, these recommendations are largely based on small retrospective studies involving selected patients [11–13]. Thus, current evidence on the safety of this approach is limited. Indeed, some experts continue to favour early surgical management for these infections [22,28]. This is reflected in a recent Delphi consensus, in which 63% of experts recommended that “An epidural empyema is almost always an indication for surgery, even in the absence of neurological deficit” [22].

A key insight from our study is the high rate of therapeutic failure among patients initially treated with antibiotics alone. The rate of failure to medical management is similar to was found in a recent systematic review [16]. Additionally, similarly high rates have been described previously in a number of works [5,6,17,29–31]. Although almost all patients in our cohort received appropriate initial antibiotics, 40% experienced therapeutic failure in the first 3 weeks, and a significant proportion required surgical rescue.

Importantly, failure of medical management can have significant consequences for patients [5,21,32]. In our cohort, patients who failed antibiotic treatment had significantly worse outcomes than those who underwent initial surgery did (frequency of primary endpoint 65.0% vs. 35.3%, p = 0.035). This worse prognosis was due mainly to higher SEA-related mortality and a higher frequency of motor sequelae. Hospital admissions were also longer in those with failure. Other recent studies have reached similar conclusions to our study [5,20,21,32]. Notably, a meta-analysis published in 2023 with involving more than 10,000 patients with spondylodiscitis (with and without SEAs) showed reduced mortality associated with early surgery compared with initial medical management [19].

Despite these data, the risks and morbidities associated with surgical intervention should not be forgotten [33–35]. In our series, we observed that the majority of patients (almost 60%) actually experienced a favorable clinical outcome with medical management. Therefore, we believe that it is of vital important to define those factors associated with higher risk of failure. In our study, we found that diabetes mellitus, an ESR greater than 75 mm/h, ventral epidural involvement and MRSA were associated with higher rates of failure. All these factors have been previously associated with medical treatment failure by other authors [5,6,28,30–32,36]. Patel et al [5] described higher rates of failure in patients with diabetes mellitus and very high acute phase reactants, whereas Spernovasilis et al [36] found that MRSA was associated with medical treatment failure. The aforementioned factors have biological plausibility that explains their association with worse outcomes. Hyperglycemia present in patients with diabetes and immunological impairments reported with this disease could favor abscess growth [37]. Likewise, a greater degree of analytical alteration (whether measured by ESR or C-reactive protein) could be associated with greater infection burden [5]. In ventral abscesses, medical treatment is likely to be offered more frequently than in posterior abscesses because of the technical difficulty to access to this location. Finally, the antibiotics commonly used in MRSA infections (vancomycin or daptomycin) have poor penetration of the blood‒brain barrier, [36,38]. Another important factor may be the level of involvement, with worse outcomes reported in patients with cervical abscesses [33,38]. In our cohort, we were unable to demonstrate this association, perhaps due to the small number of patients with cervical involvement.

Taken together, these observations support an individualized patient-centred approach. Early surgery may be considered for patients with low operative risk who present with factors associated with medical treatment failure, even when neurological deficits are absent [3,7,22,27]. On the basis of several (but not all) of the previously mentioned factors, there has been proposed a web-based application that can inform the probability of medical treatment failure using artificial intelligence-based algorithms [31]. Nevertheless, more studies are needed to refine the risk stratification. Specifically, prospective and randomized trials are needed to prove the potential benefit of early surgery in neurologically intact patients with risk factors for failure.

This study provides important insights into SEAs management, although it is not without limitations. This was a single-center retrospective study, with inherent limitations due to its design. First, our relatively small sample size may have limited the analysis of factors associated with failure. Nevertheless, our results are consistent with recent literature. Second, we could not analyze morbidities associated with surgery, especially those that were not clearly measurable (i.e., postoperative pain). However, the lower mortality and motor sequelae observed in this group support the hypothesis of its safety in selected low-risk patients. Third, we could not determine the exact timing of the neurological deficit at presentation, which could have influenced the absence of surgery indications in patients with motor neurological deficits. Yet, recent studies have shown that patients with SEAs could experience neurological improvement after surgery even with established neurological compromise (more than 24–48 hours) [39,40]. Accordingly, a longer duration of neurological deficit onset may not be a categorical contraindication to surgery in SEAs. Finally, the observational nature of our work prevents us from drawing roboust conclusions and to recommend specific treatment pathways. Yet, our study supports the need for an early risk stratification and poses the base for possible future research.

In conclusion, the failure rate of initial medical treatment for SEAs was high and could lead to worse outcomes. Factors associated with failure included diabetes mellitus, an ESR greater than 75 mm/h, MRSA isolation, and involvement of the ventral area of the epidural space. Initial surgical management could be considered in patients with these factors and low operative risk. For those medically managed patients, close clinical and radiological monitoring should be advised. Prospective randomized studies are required to better define initial management strategies of patients with SEA.