Discussion

In this cohort of 355 surgically treated spondylodiscitis patients, revision-free survival remained high, with rates of 80.9% at 3 years. However, 13.8% of patients (n = 48) required revision surgery due to long term construct failure. Neuhoff et al. reported on 31 patients with a follow-up exceeding one year and observed a revision rate due to implant failure of 11%, including posterior pedicle screw loosening (8%) and anterior cage subsidence (3%) (Neuhoff et al., 2024). At our institution, posterior instrumentation alone without discectomy is considered sufficient in most patients to achieve spinal stability and early mobilization, whereas circumferential reconstruction with corpectomy is reserved for severe vertebral body destruction and rigid kyphotic deformity. Although alternative interbody fusion techniques such as XLIF, TLIF, or PLIF may be considered in selected cases, Schatlo et al. reported no significant differences in clinical outcomes according to the surgical technique used for pyogenic spondylodiscitis(Schatlo et al., 2023).

Multivariable Cox regression identified relapse infection as the strongest independent predictor of long-term revision due to construct failure (HR 11.54, 95% CI 5.05–26.36; p < 0.001), while higher Charlson Comorbidity Component scores and elevated preoperative CRP levels were also consistently associated with increased revision risk. Although elevated preoperative CRP was independently associated with an increased risk of revision, the hazard ratio reflects the effect of each 1 mg/L increase and is therefore expected to be modest. Consequently, CRP should not be interpreted as an isolated predictor of construct failure but rather as a surrogate marker of the underlying inflammatory burden. When considered together with clinical findings, comorbidity burden, and evidence of recurrent infection, markedly elevated CRP levels may help identify patients who warrant closer postoperative surveillance.

The relapse of infection can compromise the entire bone structure and is more difficult to treat due to the presence of implants and increased resistance to antibiotics (Schomacher et al., 2014; Fayazi, 2004). It also triggers an ongoing inflammatory response that affects both soft tissue and bone. This inflammatory process can prevent proper healing and interfere with bone remodeling, which is critical for maintaining the stability of spinal implants (Shiban et al., 2020). The presence of a relapse infection often necessitates revision surgeries to remove the infected hardware, debride the infected tissue, and re-stabilize the spine (Schomacher et al., 2014; Fayazi, 2004). This is why it is very important to try to identify the pathogen before starting antibiotic treatment and to ensure that the infection is treated for a sufficient duration using all available diagnostic methods. Shiban et al. reported that screw loosening should raise a high index of suspicion for low-grade infection and demonstrated that sonication of explanted implants significantly increased bacterial yield in both groups (Shiban et al., 2020). In our current practice, all explanted screws are routinely submitted for sonication; however, this was not performed at the outset of the present study. We believe that relapse of infection may be a clinical indicator of an underlying low-grade infection. Sommer et al. documented revision surgery in 20% (14/70) of all patients who underwent surgery for spondylodiscitis. The most common reason for revision was recurrence of spondylodiscitis, followed by screw dislocation, and wound infection (Sommer et al., 2023). Fayazi and Schomacher et at. have reported that pseudarthrosis was observed in approximately 10% of patients following posterior fixation via interbody fusion for pyogenic spondylitis (Schomacher et al., 2014; Fayazi, 2004). A systematic review and meta-analysis on the influence of instrumentation type on outcomes after the surgical management of spondylodiscitis shows no significant differences in fusion rates. Fusion rates were 93.4% with titanium, 98.6% with allograft, 84.2% with autologous bone graft, and 93.9% with polyetheretherketone (PEEK). Screw loosening rates were 0.33% with titanium, 0% with allograft, 1.3% with autologous bone graft, and 8.2% with polyetheretherketone (Maddy et al., 2024). Lin et al. demonstrated that patients with multiple medical comorbidities frequently experience complications, such as infection relapse and pseudarthrosis. The risk of pseudarthrosis may be even higher than reported figures suggest, especially in the later stages of pyogenic spondylitis (Lin et al., 2012). Similarly, in our analysis, higher Charlson Comorbidity Index scores and elevated preoperative CRP levels were consistently associated with an increased risk of revision. No other evaluated variables were independently associated with revision risk. Even though Bettag et at. showed that patients with lower estimated bone mineral density (BMD) have an increased likelihood of requiring revision surgery due to implant failure, it also highlights a possible association between low BMD and implant loosening (Bettag et al., 2020). Osteoporosis was not independently associated with revision risk in our cohort. Postoperative wound infection and wound revision showed borderline associations with long-term revision in our univariate analyses. However, these effects did not reach statistical significance after adjustment and were attenuated in multivariable time-to-event models. This finding likely reflects the close temporal and clinical relationship between early postoperative wound complications and subsequent relapse infection, which emerged as the strongest predictor of long-term revision. Thus, wound infection and wound revision may represent intermediate events along the causal pathway rather than independent risk factors. The observed trends nonetheless suggest that early postoperative wound complications warrant close surveillance, as they may identify patients at increased risk for later revision through persistent or recurrent infection. From a clinical perspective, our findings may facilitate postoperative risk stratification. Patients with relapse infection, a higher Charlson Comorbidity Index, and elevated preoperative CRP levels appear to represent a subgroup at increased risk of construct failure. Accordingly, postoperative surveillance with serial inflammatory marker measurements (e.g., CRP) and CT imaging at approximately 3, 6, and 9 months after surgery may facilitate the early detection of recurrent infection and/or construct failure, particularly in patients with a higher comorbidity burden. However, this surveillance strategy requires prospective validation before it can be considered a standardized follow-up protocol.

Strengths and limitations

The main strength of the present study is the large cohort of patients undergoing instrumented surgery for pyogenic spondylodiscitis and the comprehensive evaluation of predictors associated with revision surgery due to construct failure. Nevertheless, several limitations should be acknowledged. First, the retrospective single-center design introduces the potential for selection bias and limits the generalizability of the findings. Surgical decision-making, including the choice of surgical approach, construct configuration, and timing of revision, was based on individual patient characteristics and surgeon judgment, despite generally consistent institutional treatment principles, and may therefore have introduced heterogeneity in treatment strategies. Second, although osteoporosis was assessed using CT-based Hounsfield unit measurements, standardized assessment of bone quality using dual-energy X-ray absorptiometry was not available for all patients. In addition, detailed radiographic alignment parameters, including sagittal balance, were not consistently documented and therefore could not be incorporated into the analysis. Third, despite adjustment using multivariable Cox regression, residual confounding from unmeasured factors cannot be excluded, particularly with regard to infection severity, pathogen virulence, antimicrobial adherence, nutritional status, and other clinical variables that may influence construct failure. Finally, although the statistical approach was designed to minimize overfitting, the relatively small number of revision events limited the number of variables that could be included in the multivariable model and may have reduced the ability to detect weaker associations. Prospective multicenter studies with standardized clinical and radiographic follow-up are warranted to validate these findings and further refine risk stratification for construct failure.