胫骨平台骨折术后膝关节创伤性关节炎风险预测模型的构建

Predictive model for knee traumatic arthritis secondary to tibial plateau fracture surgery

目的通过建立胫骨平台骨折术后创伤性关节炎（Knee traumatic arthritis，KTA）的风险预测模型，评估胫骨平台骨折术后KTA发生的可能性。方法收集2009年12月至2015年4月共258例于南通大学附属医院行手术治疗的胫骨平台骨折患者资料，其中男117例，女141例。258例被分为模型训练队列和模型验证队列，在模型训练队列中，采用单因素及多因素logistics回归模型和Nomograph图进行风险预测模型的构建；在模型验证队列中，采用拟合优度检验和受试者工作特征曲线（receiver operating characteristic curve，Roc）进行风险预测模型校准度和区分度的评价。结果258例中，2009-2012年训练队列共151例，2013至2015年验证队列共107例。训练队列和验证队列在性别、年龄等一般特征上差异无统计学意义。151例训练队列中11例KTA（7．3％，11／151）。训练队列中单因素和多因素Logistics回归分析均显示患者的体重指数、骨折类型、骨质疏松和骨折复位是胫骨平台骨折术后KTA发生的危险因素。以Logistics回归模型为基础绘制Nomo图，建立可视模型。107例验证队列中9例发生KTA（8．4％，9／107），拟合优度检验显示：Person检验（χ2=321．64，P=0．329）表明该模型具有较好的标准度；验证队列受试者工作特征曲线下面积为0．7551，表明其有中等区分度预测KTA的发生。结论高体重指数、不稳定骨折、骨质疏松和骨折非解剖复位是胫骨平台骨折术后发生KTA的独立风险因素；以此为基础构建的风险评估模型有较好校准度和中等区分度，可为临床胫骨平台骨折术后发生KTA的风险预测提供参考。

Objective Establish a predictive model for investigating the risk factors of Knee traumatic arthritis （KTA） secondary to tibial plateau fracture surgery. Methods 258 consecutive patients with tibial plateau fracture between December 2009 and April 2015 were collected, 117 cases were male, 141 cases were female. All the patients were divided into the training cohort （2009-2012） and the validation cohort （2013-2015）. In the training cohort, univariate and multivariate logistic regression was used to analyze the risk factors for traumatic arthritis; then Nomogram predicting model was built based on logistic regression model. In the validation cohort, goodness of fit test and ROC curve were used for testing calibration and discrimination power re- spectively. Results All the patients were divided into 151 eases for the training cohort and 107 cases for the validation cohort. In the training cohort （151 patients）, 11 patients were diagnosed as KTA with the incidence of 7.3%. The results showed that body mass index, type of fracture, osteoporosis and poor reduction were risk factors of postoperative KTA. In the validation cohort （107 patients）, 9 patients were confirmed as KTA, with the incidence of 8.4%. Goodness of fit test showed Pearson χ2was 321.64, which indicated good calibration. The area under the receiver operating characteristic curve （ROC） was 0.7551, which indicated moder- ate discrimination. Conclusion High body mass index, unstable fractures, osteoporosis and non-anatomic reduction are indepen- dent risk factors for KTA secondary to open reduction internal fixation of tibial plateau fractures. The predictive model has good calibration and moderate discrimination power in predicting KTA secondary to tibial plateau fractures.