Epigenetic Signatures Predict Pathologic Nodal Stage in Breast Cancer Patients with Estrogen Receptor-Positive, Clinically Node-Positive Disease

Background Breast cancer patients with clinically positive nodes who undergo upfront surgery are often recommended for axillary lymph node dissection (ALND), yet more than half are found to have limited nodal disease (<= 3 positive nodes, pN1) at surgery. In this study, we examined the efficiency of molecular classifiers in stratifying patients with clinically positive nodes to pN1 versus > pN1 disease. Methods We evaluated the clinical and epigenetic data of patients in The Cancer Genome Atlas with estrogen receptor-positive, human epidermal growth factor receptor 2-negative invasive ductal carcinoma who underwent ALND for node-positive disease. Patients were divided into control (pN1, <= 3 positive nodes) and case (> pN1, > 3 positive nodes) groups. Machine learning algorithms were trained on 50% of the cohort and validated on the remaining 50% to identify DNA methylation signatures that predict > pN1 disease. Clinical variables and epigenetic signatures were compared. Results Controls (n = 34) and case (n = 24) cohorts showed similar mean age (56.4 +/- 12.2 vs. 57.6 +/- 16.7 years; p = 0.77), number of nodes removed (16.1 +/- 7.3 vs. 17.5 +/- 6.2; p = 0.45), tumor grade (p = 0.76), presence of lymphovascular invasion (p = 0.18), extranodal extension (p = 0.17), tumor laterality (p = 0.89), and tumor location (p = 0.42). The mean number of positive nodes was significantly different (1.76 +/- 0.82, pN1; 8.83 +/- 5.36, > pN1; p < 0.001). Three epigenetic signatures (EpiSig14, EpiSig13, EpiSig10) based on DNA methylation patterns of the primary tumors demonstrated high accuracy in predicting > pN1 disease (area under the curve 0.98). Conclusions Epigenetic signatures have an excellent diagnostic accuracy for stratifying nodal disease in patients with clinically positive nodes. Validation of this tool is warranted and may provide an accurate and cost-effective method of identifying patients with predicted low nodal burden who could be spared the morbidity of ALND.

Automated summary

The study examined the efficiency of molecular classifiers in stratifying breast cancer patients with clinically positive nodes to pN1 versus > pN1 disease. The results showed that epigenetic signatures based on DNA methylation patterns had high accuracy in predicting > pN1 disease. This study may provide an accurate and cost-effective method of identifying patients who could be spared the morbidity of axillary lymph node dissection.