乳腺癌术后上肢淋巴水肿的风险预测可视化模型构建与验证

Construction and validation of risk predictive visualized model of upper limb lymphedema after breast cancer surgery

目的探讨乳腺癌术后上肢淋巴水肿的危险因素,构建风险预测可视化模型并验证。方法回顾性分析528例行手术治疗的乳腺癌患者的临床资料,将患者随机分为训练集352例和验证集176例。根据是否并发上肢淋巴水肿将训练集患者分为淋巴水肿组67例和非淋巴水肿组285例,采用Logistic回归分析筛选乳腺癌术后上肢淋巴水肿的危险因素,通过R软件绘制风险预测可视化模型并验证效能。结果乳腺癌患者术后上肢淋巴水肿的发生率为20.08%(106/528)。多因素Logistic回归分析结果显示,体质量指数(BMI)高、高血压、TNM分期Ⅲa期、双侧病变、腋窝淋巴结清扫水平Ⅲ级、术后放化疗均为乳腺癌术后并发上肢淋巴水肿的独立危险因素(P<0.05)。基于这6个独立危险因素构建列线图模型,该模型在训练集、验证集中的校正曲线均与理想曲线拟合良好;受试者工作特征曲线分析结果显示,列线图模型在训练集、验证集中预测术后上肢淋巴水肿的曲线下面积分别为0.950、0.886;决策曲线分析结果显示,该模型在训练集和验证集中预测术后上肢淋巴水肿的总体净获益均高于所有患者全干预或未干预。结论BMI高、高血压、TNM分期Ⅲa期、双侧病变、腋窝淋巴结清扫水平Ⅲ级、术后放化疗均为乳腺癌术后并发上肢淋巴水肿的独立危险因素,据此构建的风险预测列线图模型具有良好的预测效能。

Objective To explore the risk factors of upper limb lymphedema after breast cancer surgery and to construct and validate a risk prediction visualized model.Methods A retrospective analysis was performed on the clinical data of 528 breast cancer patients who underwent surgery.The patients were randomly divided into training set(352 cases)and validation set(176 cases).The training set patients were divided into lymphedema group(67 cases)and non-lymphedema group(285 cases)based on whether they had upper limb lymphedema.Logistic regression analysis was used to screen risk factors for upper limb lymphedema after breast cancer surgery.A risk prediction visualized model was constructed and validated using R software.Results The incidence of upper limb lymphedema after breast cancer surgery was 20.08%(106/528).Multivariate Logistic regression analysis showed that high body mass index(BMI),hypertension,TNM stageⅢa,bilateral lesions,levelⅢof axillary lymph node dissection,and postoperative radiotherapy and chemotherapy were independent risk factors for upper limb lymphedema after breast cancer surgery(P<0.05).A risk prediction nomogram model was constructed based on above six independent risk factors.The calibration curve of the model in the training set and validation set both fitted the ideal curve well;the receiver operating characteristic curve analysis showed that the area under the curve of the nomogram model in the training set and validation set for predicting upper limb lymphedema after surgery was 0.950 and 0.886,respectively;the decision curve analysis showed that the overall net benefitof the model in predicting upper limb lymphedema after surgery in the training set and validation set was higher than that of all patients receiving full intervention or no intervention.Conclusion High BMI,hypertension,stageⅢa of TNM,bilateral lesions,levelⅢof axillary lymph node dissection,and postoperative radiotherapy and chemotherapy are independent risk factors for upper limb lymphedema after breast cancer surgery.The risk prediction nomogram model constructed based on these factors has good predictive performance.