A nomogram for predicting fracture of the sandwich vertebrae after percutaneous vertebral augmentation: a multicenter study with 1-year follow-up.

Purpose: This work aimed to investigate the risk factors for fractures in sandwich vertebrae (SDV; an unfractured vertebra located between 2 cemented vertebrae) formed after percutaneous vertebral augmentation (PVA) and to construct a predictive model from this.

Methods: This study retrospectively analyzed patients who underwent PVA with the formation of SDV from July 2018 to July 2023 at Affiliated Banan Hospital of Chongqing Medical University. Patients were divided into a fracture group and a control group according to the presence or absence of fracture of the SDV during the 1-year postoperative follow-up period. Independent predictors were confirmed using the least absolute shrinkage and selection operator (LASSO) method, and the nomogram was constructed and transformed into an online calculator. The discrimination, calibration, and clinical applicability of the model were assessed by Area under the receiver operating characteristic curve (AUC), calibration curve analysis, and Decision curve analysis (DCA). Finally, the model was externally validated using data from another centre and internally validated using Bootstrap.

Results: A total of 259 patients were enrolled in this study, and 36 patients had fractures of SDV within one year. Multifactorial analyses showed that low bone mineral density (BMD) (OR = 4.264, 95% CI: 2.245-8.098, P < 0.001), number of PVA > 3 (OR = 3.703, 95% CI: 1.399-9.801, P = 0.008), lack of anti-osteoporosis (OR = 4.051, 95% CI: 1.573-10.430, P = 0.004), postoperative kyphosis angle of sandwich fracture segments (PKASFS) > 10° (OR = 8.273, 95% CI: 2.991-22.881, P < 0.001), and lumbar lordosis minus thoracic kyphosis (LL-TK) < 0° (OR = 3.701, 95% CI: 1.523-8.994, P = 0.004) were screened as independent risk factors. The AUC of the model constructed based on this was 0.881 (95% CI: 0.829-0.933). The calibration curves and DCA verified that the model had satisfactory practical consistency and clinical applicability. The externally validated AUC was 0.859 (95% CI: 0.788-0.930), validating the stability of the model.

Conclusions: BMD, number of PVA, anti-osteoporosis, PKASFS, and LL-TK are independent influencing factors for fractures in SDV within one year, and a model based on this had excellent predictive efficacy.