CK19 combined with contrast-enhanced ultrasound: A prediction model for non-sentinel lymph node involvement in early breast cancer.

Abstract

11539 Background: According to Z0011 and AMAROS trials, patients with breast cancer stage cT1~2cN0 and sentinel lymphnode (SLN) 1~2 involvement can avoid axillary lymphnode dissection (ALND). But the risk of non-sentinel lymphnode (nSLN) involvement in those early stage patients is still unclear and it is difficult to predicting the risk before surgery. Our previous study showed CK19 mRNA in peripheral blood had predicative value of nSLN involvement. Also, contrast-enhanced ultrasound (CEUS) is a new effective method exmaming axillary lymph node. We aim to establish a prediction model for nSLN involvement in early breast cancer using CK19 combined with CEUS score. Methods: We identified 119 cases diagnosed early breast cancer (stage cT1~2cN0 and 1~2 SLNs involvement as in Z0011 and AMAROS trials) from Oct 2015 to Nov 2016 in Zhejiang Cancer Hospital. The CK19 mRNA of peripheral blood by RT-PCR and CEUS score of axillary lymph nodes were acquired before surgery. We used logistic regression analysis for filtering out valuable predictive clinical parameters and establishing formulas to calculate the probability of nSLN involvement. Our model was compared with Memorial Sloan Kettering Cancer Center (MSKCC) nomogram, which is one of the most reliable and validated methods for predicting of nSLN. Results: The histological grade, CK19 and CEUS score were screened by logistic regression analysis into the formula to calculate the probability of nSLN involvement. The sensitivity, specificity, total accuracy of this model was 89.13%, 80.82% and 84.03%, respectively. The false negative rate was 10.87%. The model had high quality of consistency (Kappa 0.675, p < 0.01) and goodness of fit (likelihood-ratio test, -2log liklihood = 84.607). The area under curve (AUC) of ROC was significantly higher (P < 0.01) in our model (0.914, 95%CI, 0.863~0.965) than in MSKCC nomogram (0.563, 95%CI, 0.459~0.667). Conclusions: The prediction model based on CK19 and CEUS score has satisfying sensitivity, specificity and accuracy, more effective than MSKCC nomogram. It is a valuable model of evaluating the risk of nSLN involvement in early breast cancer before surgery, picking out the patients who can truly avoid ALND.