基于临床及超声声像图特征的乳腺癌风险预测模型

  目的  通过分析乳腺病灶超声征象及部分临床特征建立乳腺癌的风险预测模型。  方法  回顾性研究2007年7月至2009年1月于本院进行乳腺病灶切除活检术的连续性病例116例, 用多因素Logistic回归得到超声及部分临床征象(包括患者年龄、乳腺癌家族史、病灶硬度、活动度、形状、边界、方向、后方回声及钙化)中的独立危险因素, 提出乳腺癌风险预测模型, 并用受试者工作特征曲线评价模型效果。  结果  116例乳腺病灶中, 52例最终诊断为乳腺癌, 其中年龄大于50岁(OR=6.61, 95%可信区间1.07~40.72)、临床触诊质硬肿物(OR=6.56, 95%可信区间1.32~32.58)、超声声像图形态不规则(OR=19.93, 95%可信区间2.49~159.45)、边界模糊(OR=21.32, 95%可信区间1.98~230.14)、边缘成角或毛刺状(OR=31.33, 95%可信区间2.61~376.02)为乳腺癌的独立危险因素(P < 0.05)。据此建立乳腺癌风险预测模型, 该模型整体预测的准确性达96.7%。  结论  本研究建立的乳腺癌风险预测模型并提出的患乳腺癌风险独立危险因素, 在临床实践中具有较高的客观性和可操作性。

A Breast Cancer Risk Prediction Model Based on the Clinical Characteristics and Sonographic Features

  Objective  To propose a breast cancer risk prediction model by analyzing the clinical characteristics and sonographic features of breast lesions.  Methods  A total of 116 consecutive breast lesion samples obtained by biopsy in our hospital from July 2007 to January 2009 were retrospectively examined. Open biopsies were performed on each patient. The pathological results were used as the golden standard of diagnosis.Multivariate logistic regression analysis was used to identify the independent risk factors of breast cancer including age, family history of breast cancer, the hardness, mobility, shape, margin, orientation, posterior acoustic features, and calcification of the masses. The prediction model was developed and a receiver operating characteristic (ROC) curve was used to evaluate the efficacy of the prediction model.  Results  Of the 116 breast lesions examined, 52 breast lesions were diagnosed as breast cancer. The independent risk factors included the patient's age ofmore than 50 years old (OR=6.61, 95%CI 1.07-40.72), hard mass (OR=6.56, 95%CI 1.32-32.58), irregular shape (OR=19.93, 95%CI 2.49-159.45), instinct margins(OR=21.32, 95%CI 1.98-230.14) and angular or speculated margins (OR=31.33, 95%CI 2.61-376.02). The whole accuracy of this prediction model was 96.7%.  Conclusions  We developed a breast cancer risk prediction model and proposed independent risk factors, which can help predict the risk of breast cancer in clinical practices.