关键词: 多发性骨髓瘤, 椎体压缩性骨折, 骨髓浆细胞比例, 体质量指数, HU, 脊柱肿瘤脊椎不稳评分, 预测模型

Abstract: Objective To explore the construction of a risk prediction model for vertebral compression fractures in patients with multiple myeloma(MM). Method Prospectively selected 114 patients with MM admitted to Tangshan People's Hospital from April 2017 to January 2021 as the research objects.According to whether the patients had vertebral compression fractures,they were divided into the fracture group and the non-fracture group.Multi-factor analysis was used to explore the influencing factors of the risk of vertebral compression fractures in MM patients,and the nomogram prediction model was constructed,and the prediction performance of the model was tested through the C-index index and the calibration curve. Result As of May 1,2021,44 cases had vertebral compression fractures,70 cases had no vertebral compression fractures,and the incidence of vertebral compression fractures was 38.60%.There were statistically significant differences in body mass index,HU value,bone marrow plasma cell ratio,and spinal tumor spinal instability score(SINS) between the fracture group and the unfractured group(P<0.05).Multivariate logistic analysis showed that body mass index(OR=0.464,95%CI 0.293~0.735) and HU value(OR=0.966,95%CI 0.949~0.984) were protective factors for fractures,SINS(OR=1.285,95%CI 1.111~1.487),bone marrow plasma cell ratio(OR=1.078,95%CI 1.025~1.133) are risk factors for fracture(all P<0.05).The fracture risk prediction model based on the above factors predicts the consistency index(C-index) is 0.913(95%CI 0.863~0.963),and the calibration curve shows that the observation is in good agreement with the actual prediction.Conclusion The risk model of vertebral compression fracture in MM patients constructed by body mass index,HU value,bone marrow plasma cell ratio and SINS factors has high predictive value.

Key words: multiple myeloma, vertebral compression fracture, bone marrow plasma cell ratio, body mass index, HU, spinal tumor spinal instability score, predictive model