Association Between Hounsfield Units and Vertebral Body Re-collapse 3 Months After Posterior Fixation for Thoracolumbar Burst Fractures in Working-Age Patients: A Retrospective Register-Based Cohort Study

Study Population

This retrospective study included patients registered in the Swedish Fracture Register between 2015 and 2019. The Swedish Fracture Register is a nationwide quality register covering all fracture types in Sweden. Spinal fractures have been recorded since 2015 and are classified according to a modified AO Spine classification system, enabling standardized data collection.14,15 Eligible patients were of working age (18–66 years) and had sustained a single-level thoracolumbar burst fracture between T11 and L3 that was treated with posterior fixation. The surgical approach consisted of either open surgery or percutaneous pedicle screw fixation, with the choice of surgical technique left to the discretion of the treating surgeon. To optimize fixation strength, the largest possible pedicle screws relative to pedicle dimensions were selected and inserted. No posterior column osteotomies, including ponte osteotomies, were performed in this study.

Patients with coexisting fractures in the same-level pedicles or adjacent posterior structures (lamina, transverse, or spinous processes) were eligible if the posterior tension band was patent or indeterminate as determined with computed tomography (CT) and/or magnetic resonance imaging (MRI).16 Exclusion criteria included multiple spine fractures, neurological injury beyond a single-level root (spinal cord or cauda equina injury), low-energy trauma (defined as a fall from standing height or less), patients with incomplete radiological imaging postoperatively, and pathological fractures. Information regarding comorbidities, including diabetes, heart disease, and osteoporosis, as well as spine-related reoperations within 1 year, was obtained from the Swedish National Patient Register maintained by the National Board of Health and Welfare.17 Registry linkage was performed using each patient’s unique personal identification number, allowing identification of relevant comorbidities based on ICD-10 codes recorded during inpatient admissions or specialized outpatient visits.

Medical Image Analysis

Medical images were collected from the radiology departments of the participating hospitals. Patients underwent CT and/or MRI within 2 weeks of injury, followed by a postoperative CT scan and either radiographs or a CT scan at the 3-month follow-up, according to local clinical practice. CT and MRI examinations were performed in the supine position, while conventional radiographs were taken in the lateral decubitus position. All images were stored and analyzed using the institutional Picture Archiving and Communication System. Thoracolumbar CT examinations were performed using Siemens Somatom Definition Flash (Siemens AG, Germany) and Canon Aquilion Prime (Canon Medical Systems, Japan) in helical mode with 120 kVp, adaptive tube load, and a slice thickness of 0.6 to 0.7 mm. Multiplanar reconstructions were generated in 3 orientations (1.0–3.0 mm thickness).

The evaluation of kyphosis and vertebral body height assessments was conducted following the methodology outlined by Teng et al.18 Measurements included (1) kyphosis angle (Cobb angle between the superior endplate of the upper vertebra and the inferior endplate of the lower vertebra), (2) wedge angle (angle between the superior and inferior endplates of the fractured vertebra), and (3) the ratio of anterior to posterior vertebral body height (Figure 1). These parameters were assessed preoperatively, immediately postoperatively, and at 3 months postoperatively. Re-collapse was defined as an anterior vertebral height loss of at least 20%, in accordance with a previously published study by Jang et al.9 Although this threshold has not been formally validated, it has been consistently applied in prior radiographic outcome studies.9 The adjusted anterior height was calculated using the height ratio of adjacent vertebrae on immediate postoperative and 3-month postoperative images. Bone quality was assessed using Hounsfield unit (HU) values measured in the L1 vertebral body on the preoperative CT scan. Measurements were obtained using an oval region of interest placed within the trabecular bone on the mid-axial slice (Figure 2). When L1 was fractured, measurements were obtained from L2 instead.19

• Download figure
• Open in new tab
• Download PowerPoint

Figure 1

(A) Computed tomography image immediately after posterior fixation. (B) Radiographs 3 months after surgery on the thoracolumbar area. The height of the collapsed vertebral body was measured at the anterior (AH) and posterior (PH) borders. The kyphosis angle (KA) and wedge angle (WA) were also measured.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 2

Computed tomography images illustrating the measurement of the Hounsfield unit value in a series of axial views of L1 vertebrae.

The depth of pedicle screws within the vertebral body (% depth) was measured postoperatively based on previously described methods by Fujita et al.20 Measurements were performed bilaterally in the most cranial and caudal vertebrae within the fixation segment, and the mean percentage depth was calculated (Figure 3). In addition, the presence of screws inserted into the fractured vertebra was recorded. To minimize observer-related measurement error, 1 spine surgeon (F.B.) and 1 orthopedic resident (S.B.), neither of whom had been involved in the treatment of the study patients, independently evaluated the imaging studies. Interrater reliability for kyphosis angle, wedge angle, anterior/posterior vertebral height ratio, HU measurements, and percent screw depth was assessed using intraclass correlation coefficients (ICCs). Agreement for screw pull-out, assessed as a binary outcome (present/absent), was evaluated using the Cohen kappa coefficient. Finally, fractures were classified on preoperative CT scan according to the LSC and 2013 AO spine injury classification system.21 According to the AO classification, A3 fractures represent incomplete burst fractures involving 1 endplate and the posterior vertebral wall, whereas A4 fractures are complete burst fractures involving both endplates and the posterior vertebral wall. Posterior ligamentous injury was classified as B0 when the posterior tension band was intact and as B2 when there was complete disruption of the posterior ligamentous complex, with or without an associated posterior bony injury. Patients classified as B2 demonstrated partial disruption of the posterior ligamentous complex involving 1 to 3 of the 4 supporting structures (supraspinous ligament, interspinous ligament, facet joint capsules, and ligamentum flavum). These cases were coded as B2 because of clear evidence of posterior tension band failure, although overall posterior ligamentous complex integrity was considered indeterminate since not all 4 structures were injured. Patients were divided into 2 groups based on their LSC score: 68 patients had an LSC score ≥7, and 16 patients had an LSC score <7. The progression of the kyphosis angle, wedge angle, and anterior/posterior height ratio (A/P height ratio) was then compared between the 2 groups.

• Download figure
• Open in new tab
• Download PowerPoint

Figure 3

The measurement of the percentage depth of the screw insertions was defined as the ratio of the length of the screw in the vertebral body (A) to the length of the anteroposterior length of the vertebral body (B) on the axial computed tomography plan percentage (A/B × 100).

Measurement of Pull-Out of Pedicle Screw From the Vertebral Body

Pedicle screw pull-out was assessed by comparing immediate postoperative images with those obtained at the 3-month follow-up. The most cranial and caudal screws within the fixation vertebrae levels were evaluated. For cases with CT scans available at both time points, pull-out was defined as ≥2 mm screw displacement on axial CT images, in accordance with previously published criteria.22

When only lateral conventional radiographs were available at the 3-month follow-up, pull-out was assessed by comparing the radiographs with sagittal reconstructions from the immediate postoperative CT scan. To minimize observer-related bias, pull-out was independently evaluated by 2 spine surgeons, who were blinded to each other’s assessments. Cases in which the presence of pull-out could not be reliably determined were reviewed by a third reviewer. Five cases (3 radiograph-based and 2 CT-based) remained inconclusive despite this review and were excluded prior to the interrater reliability analysis. Of the included assessments, 86 were based on CT imaging and 9 on follow-up radiographs. In addition, all 9 cases assessed using follow-up radiographs were reviewed by the third reviewer because of the greater difficulty in evaluating screw pull-out using this imaging modality alone. Five cases in which reliable assessment of screw pull-out was not possible were excluded from the analysis. Two cases demonstrated severe rotational deformity on follow-up radiographs, precluding accurate evaluation; 1 case involved rod breakage; 1 had a dislodged set screw; and 1 had a pedicle screw positioned completely outside the pedicle. These cases were excluded to ensure the validity and reliability of the pull-out assessment.

Surgery

All patients underwent posterior instrumented stabilization using either an open or minimally invasive approach. Based on current evidence, pedicle screws were inserted at least 1 level above and 1 level below the fractured vertebra.5 Short-segment fixation, defined as instrumentation 1 level above and 1 level below the fracture, was used in 68 patients, whereas long-segment fixation was used in 31 patients. Intermediate pedicle screws placed in the fractured vertebra were used in 67 cases. In 1 patient, fixation was limited to a single segment. All procedures were performed within 2 weeks of injury by either a consultant orthopedic surgeon or a neurosurgeon specializing in spine surgery. Pedicle screw diameter was not standardized; however, according to routine clinical practice in Sweden, screws with a diameter of 5.0 mm or larger are most commonly used for thoracolumbar posterior fixation. Neither anterior reconstruction nor posterior transpedicular reinforcement of the fractured vertebral body was performed in our cohort.

Statistics Analysis

The Shapiro-Wilk test was used to assess distribution. Normally distributed variables were reported as mean with SD, and non-normally distributed variables were presented as median with interquartile range. Group differences for continuous variables were analyzed using the Mann-Whitney U test, while categorical variables were compared using Fisher exact test. Changes in kyphosis angle, wedge angle, and anterior/posterior vertebral height ratio between preoperative, immediate postoperative, and 3-month postoperative measurements were analyzed using paired t tests. Interrater reliability for kyphosis angle, wedge angle, anterior/posterior vertebral height ratio, HU measurements, and percent screw depth was assessed using ICCs. For screw pullout (binary outcome), interrater reliability was evaluated using Cohen kappa coefficient across all cases with available data. Reliability was interpreted according to the criteria proposed by Landis and Koch, with κ values of 0.00 to 0.20 indicating slight agreement, 0.21 to 0.40 fair agreement, 0.41 to 0.60 moderate agreement, 0.61 to 0.80 substantial agreement, and >0.80 almost perfect agreement.23

A multivariable logistic regression analysis was performed to identify factors associated with anterior vertebral height loss ≥20% at 3 months follow-up. Candidate predictors considered included HU and the LSC. Given the limited number of outcome events, the multivariable model was deliberately restricted to a parsimonious model including only 2 variables: HU and LSC. HU was selected based on its significant association in the univariate comparison, whereas LSC was included based on its established clinical relevance in prior studies. Consequently, both variables were retained in the final multivariable model.

This corresponds to an events-per-variable ratio of 8. Although this is slightly below the commonly recommended threshold of 10 events per variable, this ratio is generally considered acceptable when a limited number of carefully selected predictors are included. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated, and statistical significance was defined as P < 0.05. In a subgroup analysis of the entire cohort, progression of kyphosis angle, wedge angle, and A/P height ratio was compared between patients with LSC scores ≥7 and those with LSC scores <7. Statistical analyses were performed using SPSS for Mac (version 29.0; IBM Corp., Armonk, NY, USA).

Missing Data

Missing data were present for the following imaging variables: percentage depth (12%, 12/100), HU value (14%, 14/100), and LSC (16%, 16/100). For descriptive statistics and subgroup analyses, missing values were treated as blank entries and excluded from the respective analyses. For the multivariable logistic regression analysis, a complete-case approach was applied. Only patients with complete data for the variables included in the final model (HU and LSC) were entered into the regression analysis. This resulted in a final analytical sample of 84 patients.