基于LASSO回归构建乳腺癌患者新辅助化疗疗效的预测模型及列线图

Establishment of a Prognostic Model and Nomogram in Breast Cancer Based on LASSO

目的:探索乳腺癌新辅助化疗反应的影响因素,通过LASSO回归建立新辅助化疗疗效的临床预测模型,指导临床上诊疗方案的选择。方法:回顾性分析2020年1月~2021年12月435例乳腺癌新辅助化疗患者的临床病理资料,根据新辅助化疗后的病理反应分为有效组和无效组。采用χ²检验对两组的临床病理指标进行单因素分析;将有统计学意义的指标纳入LASSO回归分析,筛选出与疗效相关的显著变量并由此构建新辅助化疗疗效的临床预测模型。应用受试者工作特性曲线(receiver operating characteristic curve, ROC)评价该模型的预测性能。结果:LASSO回归筛选出了月经状态、BMI、肿瘤大小、ER状态、HER2状态、分子分型及四周期疗效评价七个预测变量,基于上述七个变量绘制列线图及ROC曲线,训练集和验证集的ROC曲线下面积(area under curve, AUC)分别为0.858和0.798,内部验证显示列线图具有较好的预测能力。结论:乳腺癌新辅助化疗有效的综合预测模型对新辅助化疗疗效有较好的预测能力,此模型可为指导患者个体化的治疗提供参考。

Objective: To explore the influencing factors of neoadjuvant chemotherapy for breast cancer, we establish a clinical prediction model and a nomogram based on LASSO for the efficacy of neoadju-vant chemotherapy, to guide the choice of clinical operation. Methods: By a retrospective analysis of the clinicopathological data of 435 patients with neoadjuvant chemotherapy for breast cancer who were newly treated from January 2020 to December 2021, according to the pathological response after neoadjuvant chemotherapy, they are divided into effective group and ineffective group. The χ2 test was used to conduct single-factor analysis on the clinicopathological indicators of the two groups: statistically significant indicators were included in the LASSO algorithm, LASSO regression was used to screen the predictors, and a clinical prediction model of neoadjuvant chemotherapy ef-ficacy was constructed. The receiver operating characteristic curve (ROC) was used to evaluate the predictive performance of the model. Results: LASSO regression screened seven indexes such as menstrual status, BMI, tumor size, ER status, HER2 status, molecular classification and four cycles of RECIST1.1 efficacy evaluation. Based on the above seven variables, the nomogram and ROC curve were drawn. The areas under the ROC curve (AUC) are 0.858 and 0.798 for the training and testing groups. Internal verification showed that the nomogram had high predictive ability. Conclusion: The comprehensive prediction model of effective response of breast cancer after neoadjuvant chemo-therapy has a good ability to predict the response of breast cancer after neoadjuvant chemotherapy, which can provide a reference for guiding patients’ individualized treatment.