Materials and methods

Patients

This retrospective instrumented spinal stabilization in combination with targeted antibiotic therapy. Patients treated conservatively without surgery and those with insufficient clinical documentation cohort study included patients who underwent surgical treatment for pyogenic spondylodiscitis at the University Medical Center Göttingen between 2013 and 2022. All included patients received or follow-up data were excluded from the analysis.

Preoperative assessment and infection management

Preoperative assessment included routine laboratory testing, two sets of blood cultures (aerobic and anaerobic), and computed tomography (CT) together with magnetic resonance imaging (MRI) to evaluate the extent of spondylodiscitis, vertebral bone destruction, epidural abscess formation, and spinal deformity. Osteoporosis was assessed using CT-based Hounsfield unit measurements of the lumbar vertebrae (L1–L4), with osteoporosis defined as a mean attenuation of <110 HU. Formal dual-energy X-ray absorptiometry (DXA) was not routinely available and therefore could not be used as a standardized diagnostic criterion. Intraoperatively, tissue samples were obtained for microbiological culture. In revision cases, explanted implants were additionally subjected to sonication, although this was not routinely performed during the early study period. Empirical antibiotic therapy, in accordance with current guidelines, consisted of vancomycin combined with ceftriaxone or meropenem until pathogen identification or in cases of negative blood cultures. Thereafter, antimicrobial therapy was tailored according to culture and susceptibility results. Antibiotics were generally administered intravenously for two weeks, followed by oral therapy to complete a total treatment duration of six weeks. In selected patients with resistant organisms or complicated infections, treatment was extended to up to 12 weeks following interdisciplinary discussion. Antimicrobial therapy was discontinued after clinical resolution of infection and normalization of inflammatory markers, including C-reactive protein (CRP) and leukocyte count.

Surgical indication and technique

At our institution, early surgical management is the standard treatment strategy for pyogenic spondylodiscitis once the diagnosis has been confirmed clinically, biochemically, and radiologically using CT and MRI. Immediate surgery was performed in patients presenting with neurological deficits. In patients without neurological impairment, surgical intervention was typically carried out within 72 h of diagnosis.

Surgical indications and the operative technique were determined by the extent of vertebral body destruction, spinal instability, and segmental deformity. Circumferential (360°) fusion was performed in patients with severe structural compromise, including >50% vertebral body destruction, pathological fractures, marked kyphotic deformity, or persistent instability requiring anterior column reconstruction. These procedures were preferentially performed in younger patients who were suitable candidates for a staged surgical approach. The first stage consisted of posterior transpedicular instrumentation performed under navigation or robotic guidance. Posterior fixation typically consisted of a short-segment construct (one level above and one level below the affected vertebra), with additional short pedicle screw placement into the infected vertebra whenever feasible, as anterior column reconstruction was planned to restore spinal stability. This was followed by anterior corpectomy with radical debridement. Anterior column reconstruction was achieved using a distractible vertebral body replacement cage (Obelisc™, Ulrich Medical, Ulm, Germany) with adjustable angulation (0°–15°) to restore sagittal alignment.

In contrast, isolated posterior instrumentation was performed in patients without substantial vertebral body destruction, pathological fracture, or significant deformity, in whom adequate spinal stability could be achieved without anterior reconstruction. Percutaneous pedicle screw–rod constructs were implanted under navigation or robotic guidance. In these patients, posterior fixation typically extended one level above and below the affected segment when the vertebral body was intact, and two levels above and below when a mild bone destruction was present, to provide adequate stability in the absence of anterior column reconstruction. However, the final construct length was individualized according to the extent of infection, bone quality, and intraoperative findings.

Patients presenting with neurological deficits and/or radicular symptoms additionally underwent spinal decompression via interlaminar fenestration, hemilaminectomy, or laminectomy, as appropriate. The surgical approach and instrumentation techniques remained consistent throughout the nine-year study period.

Postoperative management and follow-up

Postoperative assessments included a CT scan of the operated area to evaluate surgical outcome during the initial hospital stay. A hard brace was prescribed for 12 weeks after surgery. Clinical and radiological follow-up using MRI took place 8 weeks after surgery. A further clinical evaluation and CT scan were performed at three months and at one year postoperatively.

Outcome measures

The primary outcome was revision surgery due to construct failure following the index procedure for pyogenic spondylodiscitis. Construct failure included revision surgeries performed for screw loosening, screw or rod breakage, newly developed spinal deformity, pseudoarthrosis, and adjacent segment degeneration (Fig. 1), irrespective of whether these failures were caused by purely mechanical factors or occurred secondary to recurrent infection. Revision surgery was performed in association with recurrent symptoms such as pain and/or neurological deficits. Surgeries performed solely for wound revision or for correction of initial implant misplacement were not classified as construct failure (see Fig. 2).

Fig. 1: A: Patient with construct failure (screw loosening and dislocation after transpedicular instrumentation for spondylodiscitis), Cobb angle is approximately 24°. B: Extension of stabilization and corpectomy with cage placement. Cobb angle is now 40°. C: 3D representation after revision surgery.

Fig. 2: Patient flowchart showing study enrollment, exclusions, follow-up, and final revision cohort.

Relapse of infection was defined as the recurrence of symptoms indicative of pyogenic spondylodiscitis (e.g., back pain or fever) after an initial period of clinical recovery, in combination with radiological evidence on MRI or CT and/or elevated inflammatory laboratory parameters. Relapse infection was analyzed as a potential predictor of revision surgery.

Clinical data and statistical analysis

Data retrieval included patient characteristics as well as intraoperative and postoperative variables potentially associated with infection and revision risk. Data management was performed using Excel database (Microsoft Corp) and IBM SPSS Statistics Version 27.0 (IBM Corp, Released, 2016, IBM SPSS Statistics for Windows, Version 27.0, Armonk, NY, USA).

Time-to-event analyses were performed to evaluate factors associated with the primary outcome. Time was calculated from the date of index surgery to the date of revision surgery due to construct failure. Patients who did not experience the primary outcome were censored either at the time of death or at last follow-up, as appropriate. Kaplan–Meier survival analysis was used to estimate revision-free survival. Cox proportional hazards regression was applied to identify baseline predictors and postoperative factors associated with primary outcome. To avoid model overfitting given the limited number of outcome events, the multivariable Cox regression model was restricted to clinically relevant baseline variables selected a priori based on their availability before surgery, clinical relevance, and biological plausibility. Given the 48 long-term revision events, the number of covariates was further limited in accordance with the commonly accepted recommendation of approximately 10 events per predictor variable.

The baseline Cox model included age, Charlson Comorbidity Component (unadjusted for age), preoperative CRP, and osteoporosis. An extended Cox model was subsequently constructed and adjusted for key baseline confounders (age, Charlson Comorbidity Component, and preoperative CRP), with additional inclusion of postoperative and follow-up variables of interest, namely relapse infection and wound infection. Continuous variables are presented as mean ± standard deviation and were compared using Student's t-test. Categorical variables are reported as counts and percentages and were compared using the chi-square test or Fisher's exact test, as appropriate. A p-value < 0.05 was considered statistically significant.

Ethics statement

Ethical approval was obtained (ethical commission of University Hospital Göttingen, application number: 3/12/17), aligning with the 1964 Declaration of Helsinki and its amendments. All procedures adhered to local and institutional laws and data protection regulations.