Objective: To construct a dynamic enhanced MR based predictive model for early assessing pathological complete response (pCR) to neoadjuvant therapy in breast cancer, and to evaluate the clinical benefit of the model by using decision curve. Methods: From December 2005 to December 2007, 170 patients with breast cancer treated with neoadjuvant therapy were identified and their MR images before neoadjuvant therapy and at the end of the first cycle of neoadjuvant therapy were collected. Logistic regression model was used to detect independent factors for predicting pCR and construct the predictive model accordingly, then receiver operating characteristic (ROC) curve and decision curve were used to evaluate the predictive model. Results: ΔArea(max) and Δslope(max) were independent predictive factors for pCR, OR=0.942 (95%CI: 0.918-0.967) and 0.961 (95%CI: 0.940-0.987), respectively. The area under ROC curve (AUC) for the constructed model was 0.886 (95%CI: 0.820-0.951). Decision curve showed that in the range of the threshold probability above 0.4, the predictive model presented increased net benefit as the threshold probability increased. Conclusions: The constructed predictive model for pCR is of potential clinical value, with an AUC>0.85. Meanwhile, decision curve analysis indicates the constructed predictive model has net benefit from 3 to 8 percent in the likely range of probability threshold from 80% to 90%.
目的：建立基于动态增强磁共振成像(DCE－MRI)的乳腺癌新辅助治疗后病理完全缓解(pCR)的早期预测模型，并采用决策曲线分析来评价该模型的临床应用价值。 方法：连续收集2005年12月至2007年12月的乳腺癌术前新辅助化疗患者170例，对患者新辅助治疗前和新辅助治疗第一轮后的MR－T2图像进行测量。采用logistic回归筛选pCR的独立影响因素并构建pCR预测模型，使用受试者工作特征曲线和决策曲线对模型进行分析。 结果： ΔArea(max)和Δslope(max)是预测pCR的独立影响因素，OR值分别为0.942(95%CI：0.918～0.967)和0.961(95%CI：0.940～0.987)。建立的预测方程的受试者工作特征曲线下面积为0.886(95%CI：0.820～0.951)。决策曲线显示阈值概率>0.4时，随着阈值概率的增加，模型与全部手术相比，逐渐体现出更高的净获益。 结论：本研究建立的pCR模型具潜在的临床应用价值(曲线下面积>0.85)，且决策曲线显示与目前临床上全部患者均手术相比，应用此模型将使更多患者获益(阈值概率0.8～0.9，净获益3%～8%)。.
Keywords: Breast neoplasms; Decision making; Neoadjuvant therapy; Pathology.