Background
Pyogenic vertebral osteomyelitis (PVO) presents an escalating clinical challenge due to rising incidence, high mortality, and significant bone destruction. Objective quantification of vertebral body (VB) bone loss for assessing the disease severity and guiding therapeutic decisions is yet to be established.
Methods
We retrospectively identified patients with confirmed PVO between 2010 and 2020. Volumetric assessments of VBs were performed using 3D Slicer, and pre-infection volumes were estimated by linear regression based on adjacent, non-infected vertebrae. A “Destruction Quotient” (DQ) was calculated (measured volume/estimated original volume) to quantify VB loss. In a subgroup analysis VB bone loss was evaluated, depending on sex, spinal location and pathogen group.
Results
Thirty-one patients met the inclusion criteria for 3D volumetry (16 males, 15 females; mean age: 67.0 ± 9.2 years; mean BMI 32.4 kg/m²). In total, n = 267 VBs were segmented. Linear regression models demonstrated a high mean coefficient of determination (R²>0.95), with mean slopes of m = 2.3 (95% CI = 1.94–2.75) in males and m = 1.8 (95% CI = 1.46–2.19) in females. The mean measured volume of infected VBs (17.8 ± 9.3 cm³) was significantly lower than the estimated original volume (24.1 ± 10.5 cm³; p < 0.001). VBs at the lumbar spine experienced a median volume loss of 30%, whereas thoracic VBs showed 18% loss of volume. Female patients demonstrated a significantly higher median VB loss (32%) than males (12%; p < 0.05). No significant variation in DQs was observed among different pathogen groups, with Staphylococcus aureus being the most prevalent; however, within the Staphylococcus aureus subgroup, the measured VB volume was significantly smaller than the original estimated volume with a mean difference of 6.13 ± 4.9 cm3 (p < 0.01).
Conclusion
A 3D-volumetric approach and linear regression modeling offers an individualized method for quantifying VB destruction in PVO. Integrating automated segmentation and densitometric data may further enhance predictive accuracy and improve patient-specific treatment strategies.