Instrumentation removal after vertebral healing is considered beneficial in cases of posterior short-segment fixation without fusion [9, 10]. However, its indication is unclear because the vertebra may recollapse after implant removal. In this study, we developed an HPC to estimate the stability of the healed vertebra, which is based on the size and location of cavities within the healed vertebra. Previous biomechanical studies of cadavers and finite element models have revealed that boneless defects larger than approximately 1/3 of the vertebra may significantly affect the stability of the vertebra [16, 28]. Although Costa’s finite element models were designed to simulate lytic metastatic lesions, the vertebral bodies were modeled as homogeneous and isotropic materials, while the simulated lytic lesions were modeled as spherical holes [16]. These models did not embrace the osteolytic effects that are particularly notable in metastatic diseases; thus, their conclusions can be applied directly to patients with nonmetastatic diseases.
Accordingly, we considered HPC type I and type II, in which the vertebra has no cavity or has a cavity that is smaller than 1/3 of the vertebral body volume, to indicate stable healing. Vertebrae showing these types of healing tend to have sound structural properties and are unlikely to collapse again in the future. Hence, we suggest that instruments can be removed in cases of HPC type I and type II healing.
In cases of type III healing, the cavity is larger than 1/3 of the vertebral body volume but violates no more than one endplate. In this situation, the vertebra is considered unstable, conforming to the biomechanical models. Surgeons should use their discretion and not remove the implants hastily. The large cavity may keep healing and turn into a small cavity due to the dynamic nature of the healing process. We suggest that implants in cases of type III healing should not be removed until type II healing is achieved.
The location of the lytic lesion is another factor that affects the stability of the vertebra [16]. Transcortical lesions in particular cause a significant decrease in the strength of the vertebra [29]. Correspondingly, a vertebra with a large cavity that violates both endplates or the lateral cortical shell is considered significantly unstable. We classify this healing type as HPC type IV, and we do not recommend implant removal in this situation.
This classification is simple and easy to use in clinical practice. It has excellent intraobserver and interobserver agreement, according to our results. With this classification, approximately half of the healing vertebrae in our cohort were considered unstable. This condition is easy to ignore due to the asymptomatic character of unstable healing. Doctors should keep this in mind and perform CT or MR scans before considering implant removal.
Several preoperative parameters, including the LSC and its subscores, were associated with unstable healing in our cohort. Nevertheless, only the LSC comminution score was a significant predictor according to our adjusted logistic regression model. The LSC is a three-item scale originally described in 1994 by McCormack et al. to predict the failure of short-segment fixation for traumatic thoracolumbar burst fractures [22]. At that time, they found that patients with LSC scores >7 points were more likely to experience fixation failure. As spinal stabilization systems improved, modern posterior short-segment fixation was developed in combination with the use of intermediate screws, which was proven to have a lower implant failure rate [5, 20, 21]. Researchers further discovered that reduction can be satisfactorily achieved and maintained using this technique in patients with LSC score>7 points [4, 30]. Thereafter, the clinical importance of the LSC became controversial. Here, we show that the preoperative LSC comminution score can predict the healing pattern outcome of the vertebra. The LSC comminution score reflects the severity of the comminution of the body on reconstructed sagittal CT images [22]. Our results suggest that the size of the cavity is related to the severity of the comminution or involvement of the vertebra. Nonetheless, its predictive ability needs further validation before recommendation for clinical use.
Surprisingly, the LSC apposition score did not predict the occurrence of unstable healing with statistical significance. One possible explanation is that the procedure of reduction during surgery may reduce the displacement of the fracture by stretching the ligament. The kyphotic deformity item of the LSC did not correlate with the occurrence of unstable healing, suggesting that the degree of correction did not affect the vertebral healing process. These results may provide a better understanding of the healing process of vertebrae after burst fracture.
Several limitations of this study need to be mentioned. The cut-off value of this classification was selected based on the results of in vitro biomechanical models. The predictive efficacy of the HPC needs to be validated over the long term in patients after implant removal. The mixture of patients with AOSpine type A3 and A4 fractures and the inconsistency of the use of mono-/poly-axial screws in the nonfractured vertebrae may have caused bias in analyzing the healing pattern distribution. The small sample size of our cohort also hindered the discovery of predictive factors for unstable healing. Furthermore, cavities originating from Schmorl’s nodes present before injury or posttraumatic intraosseous disc herniation may be confused with those originating from incomplete healing. Even so, the HPC is applicable to cavities with different traumatic pathogeneses, as they exert similar effects on the structural properties.