Abstract P1-05-11: Improving the Performance of Early Breast Cancer Diagnosis by a Model Combining Breast Ultrasound with Methylation Markers in Non-Invasive Circulating Tumor DNA

Abstract Background Breast cancer is one of the most common cancers worldwide with the highest incidence among females in 2020 [1]. The application of breast ultrasound benefits the diagnosis of breast cancer at the early stage, but also leads to over-diagnosis. Patients with breast nodules detected by ultrasound at BI-RADS 4a or higher are usually advised to have a fine needle biopsy or surgery, although the PPVs of BI-RADS 4a and 4b categories are low (6% and 25%, respectively) [2]. Methods In this study, AnchorIRIS™ assay was used for analyzing the methylated status of cfDNA. Target enrichment was performed using an AnchorDx breast cancer-specific methylation panel consisting of 129,794 methylated markers. One hundred and twelve pairs of breast tissue and plasma samples (Malignant: Benign= 56: 56) and 40 leukocyte samples (Malignant: Benign = 20: 20) were used to identify reliable breast cancer-specific methylation markers with low noise background. Finally, a methylation model trained on 307 plasma samples (train set: test set = 214: 93) was selected for differentiating benign from malignant nodules, which was validated by two independent sets (Validation-1: Malignant: Benign = 42:46; Validation-2: Malignant: Benign = 62: 46).Results This methylation model exhibited a powerful performance on differentiating benign from malignant nodules with an AUC of 0.837 (95% CI: 0.757-0.918) in the test set, and maintained a stable predictive power with AUCs of 0.820 and 0.801 in two independent validation sets, respectively. In addition, this methylation model can reflect the difference in methylation signals between metastatic and non-metastatic cancers three years in advance. In contrast to ultrasound, the prediction rate of breast cancer is more accurate across the different age groups using the methylation model, especially for younger women less than 40 years old. Under the ultrasound BI-RADS 4 category, the accuracy (ACC) of the methylation model (IndVal-1, 73.02%; IndVal-2, 78.31%) is on average 22% higher than ultrasound (IndVal-1, 55.56%; IndVal-2, 51.81%). In both of the independent validation sets, the overall accuracy (78.7%) and specificity (SP) (58.7%) at the sensitivities above 95% of the combined model is greater than applying either ultrasound (ACC: 65%; SP: 26.1%) or the methylation model (ACC: 70.6%; SP: 52.2%) alone. Conclusion This methylation model has great potential for the diagnosis of early-stage breast cancer. It improves the diagnostic accuracy of the indeterminate breast nodules, which may assist in decreasing the unnecessary biopsies or surgeries of patients with benign lesions. The methylation model also has the potential in predicting metastatic and non-metastatic breast cancers that is valuable for patient surveillance and risk prediction. References: [1]. Siegel RL, Miller KD, Jemal A: Cancer statistics, 2020. CA Cancer J Clin 70:7-30, 2020 [2]. Spinelli Varella MA, Teixeira da Cruz J, Rauber A, et al: Role of BI-RADS Ultrasound Subcategories 4A to 4C in Predicting Breast Cancer. Clin Breast Cancer 18:e507-e511, 2018 Citation Format: Xianyu Zhang, Zhujia Ye, Yanling Yin, Liuhong Zeng, Jun Wang, Shan Lei, Marina Bibikova, Zhiwei Chen, Jian-Bing Fan, Da Pang. Improving the Performance of Early Breast Cancer Diagnosis by a Model Combining Breast Ultrasound with Methylation Markers in Non-Invasive Circulating Tumor DNA [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P1-05-11.