Zhang Xianyu, Yin Yanling, Ye Zhujia, Zhang Xingda, Wei Wei, Hao Yi, Zeng Liuhong, Yang Ting, Li Dalin, Wang Jun, Zhao Dezhi, Chen Yanbo, Lei Shan, Jiang Yongdong, Zhang Youxue, Xu Shouping, Nanding Abiyasi, Gong Yajie, Li Siwei, Yu Yuanyuan, Zhao Shilu, Liu Siyu, Zhao Yashuang, Chen Zhiwei, Yu Shihui, Fan Jian-Bing, Pang Da
Department of Breast Surgery, Harbin Medical University Cancer Hospital, Harbin, China.
AnchorDx Medical Co., Ltd., Guangzhou, China.
Cancer Med. 2025 Jun;14(12):e71004. doi: 10.1002/cam4.71004.
Breast ultrasonography and mammography remain predominant in breast tumor evaluations, yet they often result in false positives, particularly for tumors classified as BI-RADS 4a or those no more than 10 mm, which are not ideal for core needle biopsy (CNB). Early-stage breast cancer detection via circulating tumor DNA (ctDNA) methylation holds potential to bridge these diagnostic gaps.
We curated a breast cancer-specific panel by harnessing methylation profiles from in-house and public databases. Leveraging breast tissue-plasma-leukocyte samples, we identified breast cancer-specific markers, culminating in a 103-marker methylation model which underwent rigorous validation in two independent cohorts. To assess its performance, we compared it against the accuracy of ultrasonography, mammography, and CNB.
The 103-marker model exhibited remarkable proficiency in discerning benign from malignant breast tumors in plasma, with AUCs of 0.838, 0.838 and 0.823 in the validation set and two independent test sets, respectively. In BI-RADS 4a breast cancer, when compared to ultrasonography or mammography, the model augmented breast cancer diagnostic accuracy by 40.58% and 25.49%, separately. Retrospective analyses suggested that our model achieved a sensitivity of 66.67% (4/6) and a specificity of 80.36% (45/56) for surgical patients in the BI-RADS 4a category with tumors ≤ 10 mm, who did not undergo CNB, potentially sparing 45 benign patients from overtreatment. Notably, significant differences emerged in cancer scores between DCIS and invasive ductal carcinoma (p < 0.05). Higher cancer scores correlated with a more unfavorable prognosis (p < 0.05).
The 103-marker methylation model demonstrates impressive performance in distinguishing between malignant and benign tumors, facilitating precise early diagnosis of BC, and holds promise as a prognostic tool.
乳腺超声检查和乳腺钼靶检查在乳腺肿瘤评估中仍占主导地位，但它们常常导致假阳性结果，特别是对于分类为BI-RADS 4a的肿瘤或直径不超过10毫米的肿瘤，这些肿瘤对于粗针活检（CNB）并不理想。通过循环肿瘤DNA（ctDNA）甲基化检测早期乳腺癌有潜力弥补这些诊断差距。
我们利用来自内部和公共数据库的甲基化谱精心构建了一个乳腺癌特异性检测组。利用乳腺组织-血浆-白细胞样本，我们确定了乳腺癌特异性标志物，最终形成了一个包含103个标志物的甲基化模型，并在两个独立队列中进行了严格验证。为了评估其性能，我们将其与超声检查、乳腺钼靶检查和CNB的准确性进行了比较。
这个包含103个标志物的模型在辨别血浆中乳腺良恶性肿瘤方面表现出卓越的能力，在验证集和两个独立测试集中的曲线下面积（AUC）分别为0.838、0.838和0.823。在BI-RADS 4a乳腺癌中，与超声检查或乳腺钼靶检查相比，该模型分别将乳腺癌诊断准确性提高了40.58%和25.49%。回顾性分析表明，对于BI-RADS 4a类、肿瘤≤10毫米且未接受CNB的手术患者，我们的模型实现了66.67%（4/6）的灵敏度和80.36%（45/56）的特异性，可能使45名良性患者避免过度治疗。值得注意的是，导管原位癌（DCIS）和浸润性导管癌之间的癌症评分存在显著差异（p<0.05）。较高的癌症评分与更差的预后相关（p<0.05）。
这个包含103个标志物的甲基化模型在区分乳腺良恶性肿瘤方面表现出色，有助于乳腺癌的精确早期诊断，并有望成为一种预后工具。
