Association of clinicopathologic and molecular factors with the occurrence of positive margins in breast cancer

Abstract

Abstract Purpose To explore the association of clinicopathologic and molecular factors with the occurrence of positive margins after first surgery in breast cancer. Methods The clinical and RNA-Seq data for 951 (75 positive and 876 negative margins) primary breast cancer patients from The Cancer Genome Atlas (TCGA) were used. The role of each clinicopathologic factor for margin prediction and also their impact on survival were evaluated using logistic regression, Fisher’s exact test, and Cox proportional hazards regression models. In addition, differential expression analysis on a matched dataset (71 positive and 71 negative margins) was performed using Deseq2 and LASSO regression. Results Association studies showed that higher stage, larger tumor size (T), positive lymph nodes (N), and presence of distant metastasis (M) significantly contributed (p ≤ 0.05) to positive surgical margins. In case of surgery, lumpectomy was significantly associated with positive margin compared to mastectomy. Moreover, PAM50 Luminal A subtype had higher chance of positive margin resection compared to Basal-like subtype. Survival models demonstrated that positive margin status along with higher stage, higher TNM, and negative hormone receptor status was significant for disease progression. We also found that margin status might be a surrogate of tumor stage. In addition, 29 genes that could be potential positive margin predictors and 8 pathways were identified from molecular data analysis. Conclusion The occurrence of positive margins after surgery was associated with various clinical factors, similar to the findings reported in earlier studies. In addition, we found that the PAM50 intrinsic subtype Luminal A has more chance of obtaining positive margins compared to Basal type. As the first effort to pursue molecular understanding of the margin status, a gene panel of 29 genes including 17 protein-coding genes was also identified for potential prediction of the margin status which needs to be validated using a larger sample set.