Results

A total of 355 patients who underwent surgical treatment for spondylodiscitis were included in the study. The most common spondylodiscitis location was the lumbar spine found in 200 cases (56%), followed by the thoracic spine in 130 cases (37%) and the cervical spine in 25 cases (7%). The number of affected levels is as follows: one level in 127 cases (36%), two levels in 57 cases (16%), three levels in 60 cases (17%), and more than three levels in 113 cases (32%). The most common pathogen found in either blood or tissue cultures was Staphylococcus aureus (50%), Staphylococcus epidermidis (9%) and Escherichia coli (5%) and negative cultures (29%).

Revision surgery cohort due to construct failure

In this study, 48 patients (13.8%) underwent revision surgery due to construct failure. Baseline characteristics are presented in Table 1, and intraoperative, postoperative, and follow-up events are summarized in Table 2. The most common radiological reason for construct failure was screw dislocation in 36 patients (75%). Additionally, pseudoarthrosis along with degeneration of adjacent spinal segments was noted in 8 patients (17%), new deformities were observed in 12 patients (25%) and hardware breakage occurred in 2 patients (4%). All patients who underwent revision surgery had severe local and/or radicular pain, and the primary goal of the revision surgery was to reduce this complaint. Loosened screws were revised in 28 patients. In 22 of these patients, an alternative screw trajectory could be used, the dorsal instrumentation was extended by two levels above and below the affected segment, whereas in 6 patients the original trajectory was reused with thicker and longer screws and the dorsal instrumentation was extended by three levels cranially and caudally. In 4 cases, reinsertion of screws was not feasible because of severe bone destruction. In 2 patients, all instrumentation was removed, and antibiotic therapy was administered for 12 weeks without subsequent re-instrumentation. In both cases, clinical symptoms improved and no progression of kyphosis was observed following screw explantation. In cases with newly developed spinal deformity (n = 12), dorsal instrumentation was extended by two levels cranially and caudally, combined with corpectomy and placement of an expandable vertebral body cage in 8 patients. In the remaining 4 patients, dorsal transpedicular instrumentation was extended by three levels cranially and caudally without anterior reconstruction. In two cases of rod breakage, dorsal transpedicular instrumentation was extended by three levels above and below the affected segment, and two additional cross-connectors were placed to enhance construct stability.

Variable | No long-term revision (n = 299) | Surgery of construct failure (n = 48) | Effect size (OR, 95% CI) | P-value
Age (years), mean ± SD | 71 ± 12 | 69 ± 13 | — | 0.25
Male sex, n (%) | 168 (56%) | 27 (56%) | 1.00 (0.54–1.85) | 0.76
Obesity, n (%) | 25 (8%) | 4 (8%) | 1.00 (0.33–3.00) | 0.98
Smoking, n (%) | 28 (9%) | 5 (10%) | 1.13 (0.41–3.07) | 0.77
Drug abuse, n (%) | 15 (5%) | 4 (8%) | 1.72 (0.55–5.42) | 0.36
Hounsfield unit * (mean) | 117 ± 55 | 105 ± 43 | - | 0.085
MRSA infection, n (%) | 29 (10%) | 2 (4%) | 0.40 (0.09–1.75) | 0.17
Antibiotics prior to surgery, n (%) | 217 (73%) | 31 (65%) | 0.69 (0.36–1.31) | 0.14
Charlson Comorbidity Index, mean ± SD | 8.5 ± 2.0 | 9.2 ± 3.0 | — | 0.07
Preoperative CRP (mg/L), mean ± SD | 103.3 ± 94.0 | 117.5 ± 107.0 | — | 0.403
Cervical involvement, n (%) | 22 (7%) | 3 (6%) | 1.71 (0.61–4.83) | 0.46
Thoracic involvement, n (%) | 112 (37%) | 18 (38%) | 1.23 (0.66–2.28) | 0.66
Lumbar involvement, n (%) | 173 (58%) | 27 (56%) | 1.26 (0.68–2.37) | 0.39

Variable | No long-term revision (n = 299) | Surgery of construct failure (n = 48) | Effect size (OR, 95% CI) | P-value
Number of operated levels, mean ± SD | 2.5 ± 1.7 | 2.8 ± 1.4 | — | 0.20
Monosegmental surgery, n (%) | 132 (38%) | 13 (27%) | 0.61 (0.30–1.22) | 0.13
Spinal canal decompression, n (%) | 83 (28%) | 11 (23%) | 0.77 (0.38–1.59) | 0.33
Corpectomy/360° fusion, n (%) | 50 (17%) | 8 (16%) | 1.00 (0.44–2.26) | 0.66
Hospital stay (days), mean ± SD | 22 ± 12 | 26 ± 12 | — | 0.015
Wound infection, n (%) | 38 (13%) | 11 (23%) | 2.04 (0.96–4.34) | 0.073
Wound revision, n (%) | 32 (11%) | 10 (21%) | 2.20 (0.97–4.98) | 0.056
Durotomy, n (%) | 6 (2%) | 2 (4%) | 2.12 (0.42–10.84) | 0.66
Relapse infection, n (%) | 8 (3%) | 14 (29%) | 14.98 (5.86–38.29) | <0.001

Revision outcomes

At the 3-month follow-up after revision surgery, 40 of 48 patients (83%) continued to report persistent local and/or radicular pain. At the 6-month follow-up, persistent pain was reported by 35 of 48 patients (73%). Overall, 16 patients ultimately underwent circumferential (360°) fusion. Cage subsidence occurred in 5 of these 16 patients (31%), including two thoracic and three lumbar cases. In four patients, the cages remained stable within the posterior construct and no further intervention was required. One patient underwent additional revision because of cage dislocation associated with progressive vertebral destruction; the corpectomy was extended by one vertebral level and the posterior instrumentation was extended by two additional levels above and below the affected segment. At the 6-month follow-up, screw loosening was observed in 6 of 48 patients (13%). One patient underwent further posterior extension of the instrumentation, one patient underwent circumferential (360°) fusion, three patients underwent isolated removal of the loosened screws, and one patient was managed conservatively because neither neurological deterioration nor radiographic progression was observed. Four patients died within six months after their final revision procedure, three from multiorgan failure and one from pulmonary embolism.

Microbiological findings

In 14 revision cases (29%), all microbiological investigations remained negative during both the initial diagnostic workup and revision surgery despite repeated sampling. In another 14 patients (29%), no pathogen was identified initially; however, cultures obtained during revision surgery yielded positive results. The microorganisms isolated at revision included Escherichia coli (n = 3), Candida albicans (n = 1), Klebsiella pneumoniae (n = 2), Staphylococcus epidermidis (n = 2), Staphylococcus aureus (n = 4), Enterococcus faecium (n = 1), and Pseudomonas aeruginosa (n = 1), with polymicrobial infections identified in 4 patients. In 9 patients (19%), the causative microorganism differed between the initial diagnosis and revision surgery, suggesting reinfection or pathogen replacement. Six of these patients initially had Staphylococcus epidermidis infection; at revision, cultures yielded Pseudomonas aeruginosa in three cases, Staphylococcus aureus in two cases, and Proteus mirabilis in one case. One patient with an initial Pseudomonas aeruginosa infection also had Staphylococcus aureus isolated at revision surgery. In the remaining two patients, the initial pathogen was Enterococcus faecalis, whereas revision cultures yielded Staphylococcus aureus in one patient and Candida albicans in the other. In the remaining 11 patients (23%), pathogens identified during the initial diagnostic workup were not detected at revision surgery despite repeated microbiological investigations.

Analyses

Kaplan–Meier analysis demonstrated revision-free survival rates of 86.9%, 82.7%, and 80.9% at 1, 2, and 3 years, respectively (Fig. 3). Among the 48 patients who underwent revision surgery for construct failure, 12 (25%) required revision within the first 3 months after the index procedure, 20 (42%) between 4 and 12 months postoperatively, and 16 (33%) more than 12 months after the index procedure (up to 48 months of follow-up). The mean time to revision was 10.8 ± 15.6 months. Of these revision surgeries, 3 (6%) involved the cervical spine, 18 (38%) the thoracic spine, and 27 (56%) the lumbar spine. The anatomical distribution of revision surgeries should be interpreted in the context of the baseline distribution of instrumented spinal regions presented in Table 1.

Fig. 3: Kaplan–Meier curve illustrating long-term revision–free survival. Time is shown in months from index surgery to long-term revision or censoring (death or last follow-up).

Circumferential (360°) fusion was performed in 50 of 355 patients, with a mean of 4 ± 1.9 instrumented levels, and in 8 of 48 patients in the revision cohort, with a mean of 3.5 ± 1.5 instrumented levels. There was no significant difference in revision rates between groups (Fisher's exact test, p = 0.66).

In the baseline Cox regression model including age, Charlson Comorbidity Component, preoperative CRP, and osteoporosis, a higher Charlson comorbidity burden (HR 1.08 per point increase, 95% CI 1.00–1.16; p = 0.048) and elevated preoperative CRP (HR 1.004 per mg/L increase, 95% CI 1.001–1.007; p = 0.018) were independently associated with an increased hazard of the primary outcome. Age (HR 0.99, 95% CI 0.96–1.02; p = 0.623) and osteoporosis (HR 1.37, 95% CI 0.32–5.89; p = 0.668) were not significantly associated with revision risk (Table 3).

Variable | HR | 95% CI | p-value
Age (per year) | 0.99 | 0.96–1.02 | 0.623
Charlson Comorbidity Component (per point) | 1.08 | 1.00–1.16 | 0.048*
Preoperative CRP (per mg/L) | 1.004 | 1.001–1.007 | 0.018*
Osteoporosis (yes vs no) | 1.37 | 0.32–5.89 | 0.668

In the extended Cox model adjusted for baseline risk factors, relapse infection emerged as the strongest independent predictor of the primary outcome (HR 11.54, 95% CI 5.05–26.36; p < 0.001). Preoperative CRP (HR 1.005 per mg/L increase, 95% CI 1.002–1.009; p = 0.006) and Charlson Comorbidity Component (HR 1.10 per point increase, 95% CI 1.02–1.20; p = 0.014) remained significantly associated with increased revision hazard. Postoperative wound infection was associated with a higher hazard but did not reach statistical significance (HR 2.03, 95% CI 0.89–4.65; p = 0.094) (Table 4, Fig. 4). Sensitivity analyses were performed under extreme-case assumptions to address the eight patients with missing follow-up data. In the worst-case scenario, all eight patients were assumed to have undergone revision surgery due to construct failure, increasing the overall revision rate to 15.8%, whereas in the best-case scenario the revision rate remained 13.8%. Under both assumptions, Kaplan–Meier revision-free survival estimates changed only marginally, and the direction and statistical significance of all Cox regression predictors remained unchanged. Relapse infection consistently remained the strongest independent predictor of long-term revision due to construct failure with hazard ratios remaining above 10 and highly significant (p < 0.001), followed by preoperative CRP and Charlson Comorbidity Component, confirming the robustness of the primary findings.

Variable | HR | 95% CI | p-value
Age (per year) | 1.00 | 0.97–1.03 | 0.851
Charlson Comorbidity Component (per point) | 1.10 | 1.02–1.20 | 0.014*
Preoperative CRP (per mg/L) | 1.005 | 1.002–1.009 | 0.006*
Relapse infection (yes vs no) | 11.54 | 5.05–26.36 | <0.001*
Wound infection (yes vs no) | 2.03 | 0.89–4.65 | 0.094

Fig. 4: Forest plot showing hazard ratios and 95% confidence intervals from the extended Cox proportional hazards model for long-term revision. The model was adjusted for age, Charlson comorbidity component, and preoperative C-reactive protein, and additionally included relapse infection and wound infection.