During 1989–2000, the mortality rate of breast cancer decreased by approximately 43%, although the incidence rate increased year by year. However, the rate of decline in mortality is also decreasing(15). Early screening, diagnosis, and treatment of breast cancer play an important role. Breast cancer with lymph node metastasis is a risk factor for recurrence and long-term survival(2, 3). Therefore, early differentiation of regional lymph node characteristics is of great significance for diagnosing and treating breast cancer.
In this article, we retrospectively analyzed the B-mode ultrasound imaging characteristics of the relevant lymph node properties in patients diagnosed with breast cancer in our hospital and analyzed related risk factors, then constructed a new prediction nomogram model. Multivariate logistic regression analysis suggested that spiculated margins, mixture echo of the breast tumor and unclear lymphatic hilum structure were independent risk factors for breast cancer with lymph node metastasis. However, the hematological indicators, such as CEA, CA125 and CA153, were not independent affecting factors associated with lymph node metastasis. Precisely, LASSO method is a popular and robust high-dimensional predictive regression method(16). With the help of the LASSO logistic regression analysis, three risk factors were identified, including tumor spiculated margins, cortical thickness of lymph node, and unclear lymphatic hilum structure. For the first time, we have constructed and validated a novel prediction nomogram model by combining the characteristics of B-mode ultrasound images with pathological properties, which may provide a diagnosis and treatment basis to help evaluate the preoperative clinical in breast cancer.
Ultrasound imaging has the characteristics of high spatial resolution, which is more convenient for detecting enlarged lymph nodes than CT and/or MRI, and is the preferred way to observe the characteristics of lymph node enlargement(12, 17). MRI is even used to evaluate axillary lymph node metastasis of breast cancer preparation. However, MRI examinations take longer time and are more expensive than B-mode ultrasound(18). Relatively, B-mode ultrasound is conducive to clinical promotion due to its convenient management, high portability of equipment, low cost, as well as high spatial resolution(14). Predicting lymph node metastasis through indicators detected by B-mode ultrasound has significant advantages.
In fact, B-mode ultrasound imaging characteristics detection has been reported in the prediction model of lymph node metastasis in thyroid cancer, which plays an important role in the preoperative evaluation of thyroid cancer(19). Other similar studies include hepatocellular carcinoma(20), gastric cancer(21), colorectal cancer(22), and endometrial stromal sarcoma(23), etc. But not everyone is a prediction nomogram model study related to B-mode ultrasound examination. So far, there are few reports about the prediction model of preoperative ultrasound for lymph node metastasis of breast cancer using the LASSO analysis method(24). Although there have been studies on the use of ultrasound and/or radiological imaging to detect breast cancer lymph node metastasis(24, 25), so far there have been no research reports on the gold-standard preoperative ultrasound prediction model of breast cancer with lymph node metastasis by using the LASSO analysis method. Wang et al(24) collected the ultrasound images of 755 patients with early breast cancer and the radiomics analysis of the intertumoral and different peritumoral regions was carried out, constructed a prediction model by the LASSO analysis method. The results of this prediction model suggested that the AUC of the nomogram was 0.906 (95%CI: 0.882–0.930) and 0.922 (95%CI: 0.894–0.960) in the primary and external validation cohorts, respectively. In another retrospective study, 426 B-ultrasound images of early-stage breast cancer patients were included(25). Through LASSO logistic regression and multifactor analysis, a radiometric nomogram was constructed, including tumor size, lymph node status, and radiological characteristics. In the primary and validation cohorts, the AUCs were 0.78 and 0.71, respectively(25). Coincidentally, there was also a retrospective study that analyzed the ultrasound image characteristics of lymph nodes in 176 stage T1-2 breast cancer patients. The results suggested that the AUCs of the training and test sets were 0.900 (95% CI: 0.853–0.931) and 0.821 (95% CI: 0.769–0.868), respectively(26). Compared with our study, the tendency of these results seems to be consistent. However, the key elements selected for constructing the prediction model are different. In fact, Breast tumor spiculated margins, cortical thickness of lymph node, and unclear lymphatic hilum structure can be quickly obtained through B-mode ultrasound. According to the analysis of the results, the calibration chart and Hosmer Lemeshow test showed that the predicted value of the nomogram was not significantly different from the actual value in our study. At the same time, the clinical decision curve validated that the constructed prediction nomogram model was safe, reliable, and practical as described above. However, our prediction nomogram model is based on the characteristics of three significantly correlated indicators via B-mode ultrasound images screened by LASSO logistic regression analysis, which is simpler but more convenient to carry out.
Therefore, the new prediction nomogram model we constructed has important clinical significance for the early prediction of regional lymph node metastasis in breast cancer patients. Moreover, this predicted nomogram model may help in formulating diagnosis and treatment strategies, and benefit more patients in the future.
Although based on the B-mode ultrasound images analysis, we have known the risk factors for lymph node metastasis in breast cancer, also do a new predicted nomogram model which has a good diagnostic performance. However, there’re some also limitations in this study. The data from our single medical center and the heterogeneity of patients is also limited to residents from Jiangsu province and/or surrounding areas. Although the sample size is presented well, there is still a significant difference in disease distribution compared to the entire population in China and even around the world. Furthermore, the racial heterogeneity cannot be included in the study cohort. Therefore, multi-center cooperation may be a good help to remedy these deficiencies. Meanwhile, the predictive model still lacks clinical practice and validation in more patients, which requires a more rigorous follow-up research strategy in the future to obtain better feedback.