Logistic regression analysis of ultrasound findings in predicting the malignant and benign phyllodes tumor of breast

OBJECTIVE: To evaluate ultrasound characteristics in the prediction of malignant and benign phyllodes tumor of the breast (PTB) by using Logistic regression analysis. METHODS: 79 lesions diagnosed as PTB by pathology were analyzed retrospectively. The ultrasound features of PTB were recorded and compared between benign and malignant tumors by using single factor and multiple stepwise Logistic regression analysis. Moreover, the Logistic regression model for malignancy prediction was also established. RESULTS: There were 79 patients with PTB, including 39 benign PTBs and 40 malignant PTBs (33 borderline PTBs and 7 malignant PTBs by pathologic classification). The area under the ROC curve (AUC) of lesion size and age were 0.737 and 0.850 respectively. There were significant differences in age, lesion size, shape, internal echo, liquefaction, and blood flow between malignant and benign PTBs by using single-factor analysis (P<0.05). Age, internal echo, and liquefaction were significant features by using Logistic regression analysis. The corresponding regression equation In (p/(1 − p) = -3.676+2.919 internal echo +3.029 liquefaction +4.346 age). CONCLUSION: Internal echo, age, and liquefaction are independent ultrasound characteristics in predicting the malignancy of PTBs.