HER-2阳性乳腺癌新辅助化疗疗效影响因素分析及相关预后模型的构建

目的:分析接受新辅助化疗的人类表皮生长因子受体2（human epidermal growth factor receptor 2，HER-2）阳性的乳腺癌患者术后病理完全缓解（pathological complete response,pCR）的影响因素，并构建相关临床预测模型来预测pCR率。方法:收集中国人民解放军空军军医大学西京医院2017年10月至2021年05月收治的464例接受新辅助化疗的HER-2阳性乳腺癌患者的临床病理资料作为建模集；收集西安市第三医院2018年01月至2021年05月收治的91例接受新辅助化疗的HER-2阳性乳腺癌患者的临床病理资料作为验证集。分析对比建模集与验证集的临床病理特征，在建模集中通过Lasso Logistic回归模型分析，确立HER-2阳性乳腺癌患者新辅助化疗后pCR的独立危险因素，并构建列线图模型。在验证集中对模型进行外部验证，通过一致性指数（C-index）、受试者工作特征（receiver operating characteristic,ROC）曲线及曲线下面积（area under curve，AUC）对模型进行内部验证，通过校准曲线评估模型的准确性，并通过临床决策曲线分析（decision curve analysis,DCA）评价模型的临床获益和应用价值。结果:建模集和验证集中的临床病理特征进行比较，其中手术术式、化疗方案、激素受体（hormone receptor，HR）状态和T分期进行比较差异具有统计学意义（P＜0.05）。Lasso Logistic回归模型分析结果显示，N分期、靶向治疗方案、HR状态及临床疗效评估是HER-2阳性乳腺癌患者新辅助化疗后pCR的影响因素（P＜0.05），将这些因素纳入并构建列线图预测模型。建模集中模型的AUC=0.781（95%CI:0.734~0.827）；验证集中模型AUC=0.713（95%CI:0.635~0.859）。bootstrap法内部验证C-index＝0.744，显示模型无论在建模集还是验证集都具有良好的区分度。校准曲线显示列线图预测的生存率与实际生存率接近，建模集中Brier Score为0.019，验证集中，Brier Score为0.043，DCA显示模型的临床获益及应用价值较高。结论:列线图能准确预测HER-2阳性乳腺癌患者新辅助化疗后的pCR率，为临床的诊疗提供科学依据。

The analysis of the factors affecting pathological complete response of HER-2 positive breast cancer patients after neoadjuvant chemotherapy and the construction of related model

Objective:To analyse the factors affecting pathological complete response(pCR) of human epidermal growth factor receptor 2(HER-2) positive breast cancer patients after neoadjuvant chemotherapy,and construct a nomogram to forecast the pCR rate.Methods:The clinical and pathological data of 464 HER-2 positive patients received neoadjuvant chemotherapy in Xijing Hospital of Air Force Military Medical University from October,2017 to May,2021 were collected,which were set as modeling set.The clinical and pathological data of 91 HER-2 positive patients received neoadjuvant chemotherapy in Xi'an No.3 Hospital from January,2018 to May,2021 were collected,which were set as validation set.The clinical and pathological characteristics were compared between modeling set and validation set.In the modeling set,the independent risk factors of pCR in HER-2 positive patients after neoadjuvant chemotherapy were established by Lasso Logistic regression model analysis,and the nomogram model was constructed.External verification of the model was conducted in validation set.Internal validation of the model was conducted by Harrell's concordance index (C-index),receiver operating characteristic (ROC) analysis and area under curve(AUC).The accuracy of the model was evaluated by the calibration curve and the clinical benefits and application value of the model were evaluated by clinical decision curve analysis (DCA).Results:There were significant difference in surgical method,chemotherapy regimens,hormone receptor(HR) status and T staging between the patients in modeling set and validation set（P＜0.05).The results of analysis of Logistic regression model showed that N staging,targeted therapy regimens,HR status and clinical efficacy evaluation were independent risk factors of pCR in HER-2 positive breast cancer patients after neoadjuvant chemotherapy（P＜0.05）.Based on the above variables,the nomogram models were constructed.In modeling set,AUC=0.781（95%CI:0.734~0.827）,in validation set,AUC=0.713（95%CI:0.635~0.859）.The C-index for internal validation was 0.744.The calibration curve analysis showed that model predicted pCR rates had a good consistency with the actual observed values.In modeling set the Brier Score was 0.019 and in validation set the Brier Score was 0.043.The DCA showed that model can bring clinical benefit.Conclusion:The nomogram can accurately predict the pCR rates of HER-2 positive breast cancer patients after neoadjuvant chemotherapy and provide scientific basis for clinical diagnosis and treatment.