军事知识服务平台
Gustilo-Anderson Ⅲ型胫腓骨骨折术后保肢失败预测模型构建与验证
Construction and validation of a predictive model for limb salvage failure after surgery for Gustilo-Anderson type III tibiofibular fractures
摘要 目的:构建Gustilo-Anderson Ⅲ型胫腓骨骨折术后保肢失败预测模型,并验证其预测效能。方法:采用回顾性队列研究分析2014年1月至2024年12月苏州大学附属无锡九院收治的577例Gustilo-AndersonⅢ型胫腓骨骨折患者的临床资料,其中男346例,女231例;年龄19~72岁[(50.2±13.0)岁]。均实施急诊保肢手术。按伤后1年内结局,分为保肢成功组(465例)和保肢失败组(112例)。按7∶3比例随机分为训练集(403例,保肢失败组81例,保肢成功组322例)和验证集(174例,保肢失败组31例,保肢成功组143例)。收集2组性别、年龄、既往史(糖尿病史、心肺疾病史)、受伤环境涉及农/牧/渔/污等因素、Gustilo-Anderson分型、AO分型、胫神经断裂、受伤至手术时间、入院毁损肢体严重程度评分(MESS)、合并颅脑、胸部或其他肢体损伤及入院首次血清乳酸水平。采用单因素分析和多因素Logistic回归分析明确保肢失败的独立相关因素,并构建列线图。通过受试者工作特征(ROC)曲线和曲线下面积(AUC)评估模型区分度,以及 Hosmer-Lemeshow( H-L)检验和校准曲线评价模型校准度。通过决策曲线分析(DCA)评估模型临床适用性。 结果:单因素分析结果显示,训练集中糖尿病史、受伤环境涉及农/牧/渔/污等因素、Gustilo-Anderson ⅢC型、AO分型C型、入院MESS、入院首次血清乳酸水平在保肢失败组与保肢成功组间差异均有统计学意义( P<0.05)。多因素Logistic回归分析结果表明,糖尿病史( OR=2.48,95% CI 1.05,5.77)、受伤环境涉及农/牧/渔/污等因素( OR=3.93,95% CI 1.89,8.28)、Gustilo-Anderson ⅢC型( OR=2.90,95% CI 1.39,6.09)、入院MESS升高( OR=1.57,95% CI 1.36,1.82)和入院首次血清乳酸水平升高( OR=1.78,95% CI 1.50,2.15)是保肢失败的独立危险因素( P<0.05)。基于上述5项指标构建列线图,最终回归方程:Logit( P)=-8.094+0.908×糖尿病史+1.367×受伤环境涉及农/牧/渔/污等因素+1.064×Gustilo-Anderson ⅢC型+0.449×入院MESS+0.576×入院首次血清乳酸水平。训练集AUC为0.91(95% CI 0.88,0.94), H-L检验 χ2=4.05, P=0.853;验证集AUC为0.88(95% CI 0.81,0.96), H-L检验 χ2=13.49, P=0.096。训练集和验证集校准曲线均显示预测概率与实际概率具有较好一致性(训练集和验证集Brier评分均为0.09)。DCA结果显示,该模型的风险阈值概率在0.0~0.8范围内,具有良好的患者净获益。 结论:基于糖尿病史、受伤环境涉及农/牧/渔/污等因素、Gustilo-Anderson ⅢC型、入院MESS和入院首次血清乳酸水平构建的Gustilo-Anderson Ⅲ型胫腓骨骨折术后保肢失败预测模型具有较好的区分度和校准度,可作为急诊早期风险分层的辅助工具。
Abstract:
Objective:To construct a predictive model for postoperative limb salvage failure in patients with Gustilo-Anderson type III tibiofibular fractures and validate its predictive performance.Methods:A retrospective cohort study was conducted to analyze the clinical data of 577 patients with Gustilo-Anderson type III tibiofibular fractures who were admitted to Wuxi Ninth People′s Hospital Affiliated to Soochow University from January 2014 to December 2024, including 346 males and 231 females, aged 19-72 years [(50.2±13.0)years]. All patients underwent emergency limb salvage surgery. According to the outcome within 1 year after injury, the patients were divided into the failed limb salvage group ( n=112) and the successful limb salvage group ( n=465). The patients were randomly divided into a training set ( n=403, including 81 in the failed limb salvage group and 322 in the successful limb salvage group) and a validation set ( n=174, including 31 in the failed limb salvage group and 143 in the successful limb salvage group) at a ratio of 7∶3. The following data were collected in both groups: gender, age, past medical history (diabetes mellitus, cardiopulmonary disease), injury environment involving agricultural, animal husbandry, fishery, or pollution-related factors, Gustilo-Anderson classification, AO classification, tibial nerve rupture, time from injury to surgery, mangled extremity severity score (MESS) on admission, combined craniocerebral, thoracic, or other limb injuries, and initial serum lactate level on admission. Univariate analysis and multivariate Logistic regression analysis were used to analyze and identify independent risk factors for limb salvage failure and a nomogram was constructed. The discrimination of the model was evaluated by receiver operating characteristic (ROC) curve and area under the curve (AUC), while the calibration of the model was assessed by the Hosmer-Lemeshow ( H-L) test and calibration curve. Decision curve analysis (DCA) was used to evaluate the clinical applicability of the model. Results:Univariate analysis showed statistically significant differences between the successful and failed limb salvage groups in the training set in terms of diabetes mellitus, injury environment involving agricultural, animal husbandry, fishery, or pollution-related factors, Gustilo-Anderson type IIIC, AO type C, MESS on admission, and the initial serum lactate level on admission ( P<0.05). Multivariable Logistic regression analysis showed that diabetes mellitus ( OR=2.48, 95% CI 1.05, 5.77), injury environment involving agricultural, animal husbandry, fishery, or pollution-related factors ( OR=3.93, 95% CI 1.89, 8.28), Gustilo-Anderson type IIIC ( OR=2.90, 95% CI 1.39, 6.09), increased MESS on admission ( OR=1.57, 95% CI 1.36, 1.82), and an increased first serum lactate level on admission ( OR=1.78, 95% CI 1.50, 2.15) were independent risk factors for limb salvage failure ( P<0.05). A nomogram was constructed based on the above-mentioned five predictors, with the final regression equation as follows: Logit( P)=-8.094+0.908×diabetes mellitus+1.367×injury environment involving agricultural, animal husbandry, fishery, or pollution-related factors+1.064×Gustilo-Anderson type IIIC+0.449×MESS on admission+0.576×initial serum lactate level on admission. The AUC was 0.91(95% CI 0.88, 0.94) in the training set, with a H-L test result of P=0.853 ( χ2=4.05), and was 0.88(95% CI 0.81, 0.96) in the validation set, with a H-L test result of P=0.096 ( χ2=13.49). Calibration curves for both the training and validation sets showed good agreement between the predicted and observed probabilities, with Brier scores of 0.09 in both sets. DCA showed that the model provided a favorable net benefit across threshold probabilities of 0.0-0.8. Conclusion:The predictive model for postoperative limb salvage failure in patients with Gustilo-Anderson type IIIC tibiofibular fractures, based on diabetes mellitus, injury environment involving agricultural, animal husbandry, fishery, or pollution-related factors, Gustilo-Anderson type IIIC, MESS on admission, and initial serum lactate level on admission, demonstrates good discrimination and calibration and may serve as an adjunctive tool for early risk stratification in the emergency setting.
刊 名 中华创伤杂志 ISTICPKU
年,卷(期): 2026, 42(10)
分类号: R641
在线出版日期: 2026年10月15日