| Description |
BACKGROUND.There is an unlimited demand for the differentiation between benign and malignant breast tumors. Breast ultrasound and mammography are the most common examinations to evaluate breast tumors, but they still have high false positive rate, especially for tumors diagnosed as Breast Imaging-Reporting and Data System (BI-RADS) subcategory 4a, and the ones less than 10mm that are unsuitable for core needle biopsy (CNB). The methylation detection using circulating tumor DNA (ctDNA) has an ability to detect breast cancer (BC) in early stage, as well as a potential to compensate for the above-mentioned defects.METHOD. A BC-specific panel was developed using the methylation profiles from in-house (338 breast tissue samples, Malignant: Benign =283:55) and public databases. 112 paired breast tissue-plasma samples (Malignant: benign=56:56) and 40 paired breast tissue-plasma-leukocyte samples (Malignant: benign=20:20) were used for screening BC-specific markers. A 103-marker methylation model was built by a model development cohort including 307 plasma samples with breast lesions (Malignant: Benign=154:153) and it was further validated in two independent test cohorts (Test-1, Malignant: Benign=42:46; Test-2, Malignant: Benign=60:46). We evaluated its performance by comparing it with ultrasound, mammography, and CNB.RESULTS. The 103-marker methylation model showed a great performance on differentiating benign and malignant breast tumor using plasma with AUCs of 0.838, 0.838 and 0.823 in the validation sets and two independent test sets, respectively. Compared to using ultrasound or mammography alone, it improved the accuracy of BC diagnosis by 40.58% and 25.49%, separately. Through a retrospective analysis, the methylation model would increase the true positive rate by 17% of patients in the BI-RADS 4a who underwent surgery due to tumor size less than 10mm that could not be punctured by core needle. In addition, it would elevate the true negative rate by 20% of patients in BI-RADS 4a and above categories whose tumor size were larger than 10mm that were diagnosed as benign lesions by CNB. According to the cancer score predicted by the methylation model, there is a significant difference between ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (p<0.05). Survival analysis showed that higher cancer scores were associated with worse prognosis (p<0.05).CONCLUSION. The 103-marker methylation model possessed a powerful performance on distinguishing breast cancer from benign tumors as well as a high accuracy in the early diagnosis of BC. It could make up for the defects of ultrasound or mammography, especially for tumor diagnosed in BI-RADS 4a category. Moreover, it could be a compensative approach for the tumors which were not suitable detected by CNB in order to decrease overtreatment rate. |
