Introduction

Among the possible causes of thoracolumbar pain, infection may be included in the differential diagnosis when clinical and/or laboratory values are suggestive. Other times, patients with a spine infection present for advanced diagnostic imaging despite infection not being clinically suspected. Several imaging findings have been associated with septic facet arthritis and discitis osteomyelitis (DOM). Specifically, endplate erosion, T2 hyperintense disc signal, disc enhancement, epidural phlegmon/abscess and paraspinal abscess have been described as suggestive of pyogenic DOM [1,2,3]. Meanwhile, other imaging features have been associated with non-pyogenic infectious causes such as the relative sparing of disc spaces and subligamentous spread of fluid attributed to Potts disease [4, 5]. However, especially early in the disease course, non-infectious processes such as degenerative disc disease, facet arthropathy and dialysis-related spondyloarthropathy can mimic the imaging appearance of DOM [3, 6]. Conversely, there are imaging findings that have been reported to reduce the likelihood of infection. Gas within the disc space and a lack of paraspinal inflammation have been reported to be negative predictors of DOM [3, 6].

Often, when radiology reports suggest the possibility of DOM or septic facet joints, percutaneous biopsy is requested to confirm the diagnosis and to guide antimicrobial therapy. It has been reported that certain organisms are more prone to causing DOM with Staphylococcus species commonly implicated [7, 8]. Other organisms such as tuberculosis and less pathogenic microbes such as Cutibacterium acnes (previously referred to as Propionibacterium) have been reported to cause spine infections with some authors suggesting that Cutibacterium may be associated with type 1 Modic endplate changes [9,10,11].

Typically, these invasive procedures require use of a CT suite, a CT technologist to run the machine, nursing staff to assist with pre-procedural and post-procedural care, additional nursing staff to assist with moderate sedation and a radiologist to perform the biopsy. Although these biopsies are frequently requested, there are limited data as to which patients would be most likely obtain a positive biopsy culture result. This is important as one of the primary objectives of the biopsy is to identify the causative organism and to tailor the antimicrobial therapy regimen. The purpose of this study is to identify associations between specific qualitative cross sectional imaging findings, specific high risk medical comorbidities and laboratory values and a positive percutaneous biopsy outcome.

Materials and methods

In this retrospective IRB-approved HIPAA complaint study (consent waived), consecutive patients presenting over a three-and-a-half-year period to a tertiary care system in the United States for suspected thoracolumbar spine DOM and/or septic facet arthritis who underwent percutaneous biopsy were identified. Spine biopsies for known or suspected tumor were excluded. For each patient, a chart review was performed and the following information was extracted (as available): age and gender, anatomic location of the biopsy (thoracic or lumbar spine and disc level), antimicrobial administration and duration, biopsy and blood culture results (including antibiotic susceptibility), anatomic pathology from the biopsies and laboratory values including white blood cell count (WBC), erythrocyte sedimentation rate (ESR), c-reactive protein (CRP), hemoglobin A1c (HbA1c) and serum creatinine (Cr). Additionally, a history of diabetes mellitus (DM), injection drug use (IDU), prior spine surgery near the biopsy site, end-stage renal disease (ESRD) and use of hemodialysis was recorded. One musculoskeletal fellowship trained radiologist (9 years’ experience) blinded to the clinical and laboratory data, reviewed the magnetic resonance imaging (MRI) and diagnostic reports and the following qualitative items were recorded: presence of a paraspinal fluid collection/abscess (including psoas and lateral and posterior paravertebral), whether the interpreting radiologist noted suspected septic facet arthritis, the presence of any disc enhancement, facet joint effusion or erosion, vacuum disc phenomenon (noted on computerized tomography, CT), edema (on T2FS or STIR) or enhancement near the vertebral body, posterior paraspinal edema or enhancement near the facet joints, disc T2 signal (greater than or equal to cerebrospinal fluid, CSF), and the presence of an epidural abscess. All paraspinal collections / epidural abscesses initially described on the clinical radiology reports were concordant with what the blinded radiologist found. A biopsy result was considered positive if the microbiology noted the presence of a microbial infection or if the surgical pathology indicated the presence of acute discitis osteomyelitis. For paraspinal fluid collection aspirations, these were done in lieu of targeting the disc/endplates.

Diagnostic performance for specific imaging features and for blood cultures were evaluated. Categorical variables were compared by Pearson’s chi-squared test or Fisher’s exact test. Continuous variables were summarized by frequency and percentage and evaluated by Wilcoxon rank-sum test. A Cohen’s kappa coefficient (κ) was calculated to measure the agreement between biopsy culture results and biopsy anatomic pathology results. Statistical analysis was performed on SAS version 9.4 software (SAS Institute, Inc., Cary, North Carolina) and a p < 0.05 was considered statistically significant.

Results

Overall outcome of percutaneous biopsies

121 patients were enrolled among which 43 biopsies were considered positive (35.5%). 38 of these (31.4%) included a positive biopsy culture while 5 were from biopsy surgical pathology alone. 72 of the 121 (60%) had both biopsy culture and anatomic pathology available while the remaining 49 having either culture or anatomic pathology only.

Outcome of microbiology and anatomic pathology assessment and agreement between methods

Among biopsies with both culture and anatomic results available, 9 (13%) had a positive culture and positive pathology, 7 (10%) had a positive culture and negative pathology, 4 (6%) had a negative culture and positive pathology while both were negative in 52 (72%). There was moderate agreement between biopsy culture and pathology (κ = 0.53, SE 0.12, 95% CI 0.28–0.77).

Association between suspected septic facet joint aspirations and microbiologic/pathologic biopsy outcome

Among the cohort, there were 7 who had suspected facet joint sepsis that had an aspiration (no bone biopsies). Among these, 2/7 resulted in a positive microbiology culture (not statistically significant).

Repeat biopsy outcomes

Five patients in our cohort had repeat biopsies. Four of them were repeated within 1–2 months after an initial negative biopsy among which 50% returned with positive results. One patient had an initially positive biopsy and a subsequent negative biopsy 5 months after the initiation of antimicrobial therapy. These repeat biopsies were not counted in statistical analyses and are reported to provide relevant post-biopsy care information only.

Association between comorbidities and laboratory values and microbiologic/pathologic biopsy outcome

Regarding patient demographic and comorbidities, biopsy yield was not associated with age, gender, DM, IDU, ESRD, use of hemodialysis, and prior nearby spine surgery (Table 1). There was no statistically significant correlation between biopsy yield and WBC, ESR, CRP, Cr or HbA1c (Table 2). Blood cultures and biopsy cultures were available for 113 (93%) of the cohort. For those with blood culture results, there was a statistically significant difference between biopsy results in those with positive blood cultures and those with negative blood cultures (p = 0.03, Table 2). Among 30 patients with a positive blood culture, 47% had positive biopsy results, while in those with a negative blood culture, results were positive in only 24% of biopsies. In the 14 cases where there were positive blood cultures and positive biopsy cultures, 93% (13/14) had concordant antimicrobial susceptibility profiles. Twelve of those thirteen were exact pathogen matched between the blood and biopsy cultures. Among the 13 cases with concordant antimicrobial susceptibility profiles, Staphylococcus aureus was the most common bacteria (n = 5). Regarding the case with positive blood and biopsy cultures, but with a discordant result and antimicrobial susceptibility profile, the blood culture was positive for Staphylococcus simulans while the spine biopsy culture was positive for Candida parapsilosis. Among these 14 patients, the mean time interval between blood culture results and biopsy results was 22.5 days (range 0–175). However, 12/14 of these patients had a mean of 2 days (range 0–5). There were two outliers of 116 and 175 day intervals.

Table 1 Demographic and clinical information. Rates of positive and negative spine biopsy results based on age, gender and the presence or absence of specific risk factors based on clinical history. For spine surgery, this is had a previous spine surgery near the level of biopsy
Full size table
Table 2 Correlation of laboratory data and biopsy results. 94.2% of patients had blood culture data. *Only blood cultures correlated with biopsy results with statistical significance. Of note, this calculation was done for those with blood culture data and biopsy culture data. Those with anatomic pathology data alone were excluded for this analysis. Among the 5 considered positive by anatomic pathology only (missing or negative biopsy culture), 80% had negative blood cultures
Full size table

Of the 38 biopsies with positive biopsy cultures, 42 organisms were recovered (including two with polymicrobial result) with Staphylococcus species being the most common genus (42.9%, Fig. 1). The administration of pre-biopsy antibiotics was correlated with biopsy results. 26/55 (47%) patients who received antibiotics had positive biopsy results while 17/66 (26%) patients who did not receive antibiotics had positive biopsy results (p = 0.014). However, among those who received antibiotics, there was no correlation between the average number of weeks of pre-biopsy antibiotic administration and a positive biopsy result (negative results: 3.5 + 3.4 weeks versus positive results: 2.7 + 4.3 weeks, p = 0.12).

Fig. 1
figure 1

Organisms recovered from percutaneous spine biopsies. 42.9% were Staphylococcus species with 11 being Staph. aureus, among which 8 were MRSA. The next most common non-staphylococcal organism was Pseudomonas aeruginosa accounting for 9.5% of recovered organisms. Please note, some cases recovered more than one organism.

Full size image

Association between imaging findings and microbiologic/pathologic biopsy outcome

Most of the imaging findings analyzed did not have a significant correlation with biopsy yield. Specifically, there was no correlation between biopsy results and the presence of disc enhancement, facet joint effusion or erosion, whether the interpreting radiologist questioned or diagnosed septic facet arthritis, disc T2 hyperintense signal, or the presence of paraspinal edema or enhancement near the vertebral body or posterior elements. There was, however, a statistically significant correlation between biopsy results and the presence of a paraspinal fluid collection or the presence of an epidural abscess (Table 3). 65% of those with paraspinal fluid collections had a positive biopsy yield while 25% without fluid collections had a positive biopsy yield (p = 0.003). There was also an association between the presence of an epidural abscess and biopsy yield (p = 0.018). When comparing sampling of the bone/disc to sampling of a paraspinal fluid collection (when present), there was a significant correlation with positive results in fluid collection sampling (positive in 71% of cases with fluid collections compared to 31% of cases without fluid collections where bone/disc sampling was performed, p = 0.006). Among the 23 cases where vacuum disc was noted, 22% had positive biopsies (p = 0.34, Fig. 2-4).

Table 3 Imaging findings. In some cases, contrast was not given or MRI was not available for analysis. For paraspinal fluid collections, all thoracic spine biopsies were excluded as they were above the psoas muscles and no posterior thoracic level collections were present. Therefore, some imaging findings will not reach the 121 patient total. *Among the imaging findings, paraspinal fluid collections and epidural abscesses were the only ones which correlated with biopsy yields with statistical significance. Sampling from a fluid collection also statistically correlated with biopsy yield. ^T2 signal was considered hyperintense if it was either equal to or greater than CSF
Full size table
Fig. 2
figure 2

58-year-old male with biopsy proven L3/L4 DOM due to MRSA. CT immediately prior to starting the procedure shows gas in the disc space with surrounding endplate erosion. On T1 there is low T1 with endplate erosion while on T2FS, there is high signal in the endplates and erosive change. The disc is similar in signal to the other non-affected discs

Full size image
Fig. 3
figure 3

60-year-old male with biopsy proven L4/L5 DOM (acute inflammation and polymorphonuclear leukocytes on surgical pathology, microbiology specimen not sent). CT immediately prior to starting the procedure shows gas in the disc space. On T1 there is low T1 signal while on T2, there is fluid signal in the disc. The endplates are intact on CT

Full size image
Fig. 4
figure 4

75-year-old male with Pseudomonas aeruginosa DOM L4/L5 and L5/S1. T1 shows low signal around the L4/L5 and L5/S1 disc spaces. There is relatively minimal endplate edema compared to the T1. Both show erosive changes at the endplates. The contrast enhanced T1FS shows marked endplate and vertebral body enhancement with some enhancement of the disc. There is also periosseous enhancement and epidural enhancement/phlegmon. On the CT, a tiny bubble of gas is seen in the anterior L4/L5 disc space

Full size image

Diagnostic performance

Diagnostic performance calculations were made for imaging findings and blood cultures. The presence of paraspinal collections and epidural abscesses were specific, but not sensitive (Table 4). Positive blood culture results were also more specific (80.0%) than sensitive (35.9%).

Table 4 Diagnostic performance for selected radiologic and lab values. (95% confidence intervals). These results use both biopsy culture and biopsy histopathologic result as the reference standard
Full size table

Discussion

In this study, the presence of specific comorbidities, laboratory and imaging data were analyzed to see whether they could predict which patients with suspected spine infections would be likely to have a positive biopsy result. Targeted antimicrobial therapy could reduce morbidity and cost associated with empiric broad-spectrum antimicrobial coverage [12]. It is also important to note that prolonged outpatient IV administration of antibiotics has associated risks [13]. Patients can potentially be managed with outpatient oral antimicrobial therapy to avoid some of these risks. However, even these oral medications are associated with potential side effects, cost to the patient and may contribute to the emergence of antimicrobial resistant organisms. A secondary biopsy objective would be to refute the diagnosis of infection and avoid the use of antimicrobial therapy altogether. If the clinical provider believes that the likelihood of an underlying infection remains, additional management including empiric antimicrobial therapy and repeat biopsy may be necessary.

In this cohort, none of the laboratory values commonly used to assess for the likelihood of DOM correlated with a positive biopsy yield. Additionally, clinical factors such as DM, previous regional spine surgery, hemodialysis, and IDU were also not associated with a positive biopsy yield. The majority of imaging findings, while potentially indicating the presence of infection, also were not associated with a positive biopsy yield. This is not to say that CT and MRI are not accurate in making the diagnosis of DOM nor that the knowledge of comorbidities and lab values are not helpful. This study did not attempt to assess for those outcomes. Instead, the association of these factors with a positive biopsy yield was assessed. For this purpose, the utility of imaging was in the detection of fluid collections as they correlated with a positive biopsy outcome. Moreover, targeting paraspinal fluid collections, when present, was preferable to sampling the intervertebral disc or adjacent endplate as it was associated with an increased correlation to a positive biopsy result. In general, targeting paraspinal fluid collections is a technically easier procedure that requires neither bone drilling nor an incision (Fig. 5). A fluid collection needle aspiration is more feasible to do with local anesthesia alone compared to bone drilling. Moreover, although this was not specifically tested, a single needle aspiration of a fluid collection would presumably be faster than a procedure requiring the use of a coaxial bone biopsy system.

Fig. 5
figure 5

50-year-old female with L4/L5 MSSA related DOM and septic facet arthritis. T2 shows fluid in the disc space and the left facet joint herniating posterior into the paraspinal muscles. T1FS + contrast shows rim enhancement of the fluid in the disc and related to the infected left facet joint. There is extensive paraspinal enhancement. CT shows needle placement in the fluid from the left facet joint

Full size image

Interestingly, although there was an association with antibiotic administration and biopsy results, the opposite of what was expected was observed. In this cohort, those who received antibiotics were more likely to have a positive biopsy result. There was no statistically significant difference in duration of antibiotic administration between those who had positive and negative biopsy results. One theory to explain this observation is that patients who are more symptomatic and may warrant urgent empiric antibiotics (e.g., sepsis) could have a higher pathogen burden and thus remain culture-positive. These results suggest use of empiric antibiotics (when needed) is not a contra-indication to percutaneous biopsy.

Among those with positive biopsy and blood cultures, the level of agreement was high. These results suggest that a positive blood culture would be sufficient to negate the need for percutaneous biopsy (especially of bone) prior to antimicrobial therapy. The Infectious Diseases Society of America guidelines for native vertebral osteomyelitis recommend against image-guided biopsy among patients with DOM and positive blood cultures for S. aureus, S. lugdunensis, or Brucella spp. [14]. In this cohort, several patients had matching blood and biopsy cultures for S. aureus. Although guidelines suggest these positive blood cultures would have been sufficient, biopsies may be performed before blood culture results are available. It may be impractical to wait three to five days for final blood culture results before performing the biopsy.

There was moderate correlation between the microbiologic and the histopathologic results from the biopsies. Discrepancies between microbiology and surgical pathology in the diagnosis of infection has been documented [15, 16]. As the primary goal of the biopsy is to direct antimicrobial therapy, the goal of a percutaneous biopsy should be for microbiological analysis. In the event that only limited biopsy material can be obtained, preference should be given to microbiology over surgical pathology. However, if the pre-biopsy differential diagnosis is infection and neoplasm, it would be important to make every attempt to get enough material to submit for both.

Our cohort was concordant with several other studies which describe organisms that commonly are associated with spine infections [7, 8, 17]. Staphylococcus aureus and coagulase-negative Staphylococcus were the most common organisms isolated from our biopsies with Staphylococcus species accounting for 43% of cases. The next most common were Pseudomonas, Escherichia coli and Cutibacterium acnes.

There are several limitations associated with this study. First, this is a retrospective study and we could not control for factors including the biopsy device utilized, multiple medical comorbidities that were not considered in this analysis, and the multiple providers that performed the biopsies. While we did not control for the biopsy device that was utilized, we did separate patients into those that had paraspinal fluid collection sampling and bone/disc sampling. Regarding the medical comorbidities, we assessed for commonly implicated risk factors associated with spine infections and those associated with DOM mimicry on diagnostic imaging. Another important limitation that should be noted is that a negative biopsy does not exclude infection. However, it would not be feasible to have every patient have open surgery with removal of all potentially infected tissue so we have to use the biopsy results as our gold standard knowing that some of these patients may still have infections despite negative biopsies. There were four patients with repeated biopsies after an initial negative biopsy. Among those, 2 returned with positive results. Future investigations into the utility of second biopsies among a larger cohort would be beneficial to guide management of this cohort. Next, patients with suspected neoplasm and those with suspected cervical spine infections were not included in this study. Another limitation is the time delay between blood culture results and biopsy results. While there were two outliers with larger time intervals, the majority of patients with both blood and biopsy cultures were biopsied in a more narrow timeframe (0–5 days). This makes a new confounding infection during that time period less likely. Finally, defining “prior antimicrobial therapy” leading up to the biopsy was challenging. Every effort was made to extract information about what type of therapy was prescribed, the length of therapy duration and patient compliance. However, it should be noted that despite best efforts, the retrospective nature of this study makes these data limited and they should be interpreted with caution.

Conclusion

In summary, this study highlights several key factors in the management of patients with suspected thoracolumbar spine infections. First, an extensive laboratory workup is unlikely to predict those that would be more likely to have a positive percutaneous biopsy result. The most useful blood test that can be obtained is the blood culture. Direction of antimicrobial therapy based on data obtained from positive blood cultures covered for organisms recovered by biopsy in the majority of cases. In the event of a positive blood culture, based on the data from this cohort, a percutaneous biopsy may not provide additional useful information. The most important imaging findings to identify would be the presence of paraspinal fluid collections as these correlated with positive biopsies. When a paraspinal fluid collection is identified, they should be targeted preferentially compared to the intervertebral disc and vertebral body endplates as aspirations yielded positive results at a higher rate than disc/endplate biopsies and such a procedure is less invasive and likely faster. Although the commonly described MRI imaging findings for DOM should be identified and used in suggesting the diagnosis, their presence alone is not associated with a positive biopsy result. While vacuum disc phenomenon has been described in association with degenerative spondylosis and has been reported to decrease the likelihood of a spine infection, in our cohort, we did find patients who had disc gas in the presence of infection. This imaging finding alone does not exclude infection and should not discourage a percutaneous biopsy. Finally, previous antibiotic administration did not reduce the biopsy yield in this cohort. Therefore, prior antibiotic administration should not be a contraindication to a percutaneous biopsy.