| Background and Objective: Evidence-based medicine Class I evidence recommends the use of sentinel lymph node biopsy(SLNB)for axillary staging in patients with negative axillary lymph nodes in clinical examination,pathological biopsy confirmed invasive breast cancer(c T1-2N0M0)or carcinoma in situ with micro-invasive.Although SLNB has a lower incidence of complications than axillary lymph node dissection(ALND),there are still similar complications,such as lymphedema,nerve damage,seroma,etc.In order to avoid axillary surgery and its complications,the European Institute of Oncology of Milan conducted a prospective multi-center randomized controlled clinical trial(SOUND clinical trial)to explore whether preoperative ultrasound can be used to non-invasively assess axillary lymph nodes.It is still in the follow-up phase.If the SLN status of patients can be distinguished non-invasively before surgery,SLNB can be avoided for SLN-negative patients,or axillary radiotherapy for SLN-positive patients,then the axillary surgery can be avoided.Therefore,predicting SLN status has important clinical significance.Ultrasonography is suitable for the evaluation of superficial lymph nodes,but it is easily influenced by the adipose tissue inside the axillary.However,axillary magnetic resonance image can(MRI)evaluate deep axillary lymph nodes without being affected by adipose tissue.Therefore,this study analyzed the relationship between preoperative axillary ultrasound examination,MRI examination,preoperative pathological features of primary breast tumors and postoperative pathology of SLNB in biopsy proven carcinoma in situ with micro-invasive and invasive breast cancer patients(c T1-2N0M0),in order to screen out the characteristic indexes most relevant to SLN metastasis among the results of axillary ultrasound examination,magnetic resonance examination and preoperative pathological biopsy,and use the found characteristic indexes to establish preoperative prediction of sentinel lymph node metastasis in c T1-2N0M0 breast cancer patients risk nomogram model to explore the feasibility of diagnosis of sentinel lymph node status through preoperative axillary ultrasound examination,magnetic resonance examination and pathological features of the primary tumor.Methods: Retrospectively collected from 2018-7-1 to 2020-7-1 in the First Affiliated Hospital of Guangxi Medical University through vacuum biopsy or local breast tumor biopsy and pathologically confirmed that it is carcinoma in situ with micro-invasive or invasive breast cancer patients(c T1-2N0M0)who have undergone breast axillary ultrasound,axillary MRI,and SLNB,a total of465 patients were enrolled according to the inclusion/exclusion criteria,and the patients were randomly divided into the model group(n=325)and The verification group(n=140)at a ratio of 7:3 using R software.The group patients were divided into SLN-positive group and SLN-negative group according to the pathological results of SLNB.Collected clinical data(age,tumor location,tumor T stage)of 465 patients,preoperative clinical pathological biopsy results(pathological type,histological nuclear grade,ER,PR,Her-2,Ki-67,lymphovascular invasion,perineural invasion),axillary ultrasound(lymph node long/short diameter,lymph node cortex,lymph node hilum,lymph node blood flow pattern)and axillary lymph node MRI(lymph node hilum,lymph node edge,lymph node cortex,bilateral lymph nodes are symmetrical),using SPSS20.0,chi-square test was performed on the count data and t test was performed on the measurement data between the model group and the verification group to analyze whether the data collected between the two groups were statistically different.Then use univariate logistics regression in model group between SLN positive group and SLN negative group to screen out the variables related to SLN metastasis.After the variables are found,collinearity diagnosis is used to prove that there is no collinearity,and then multivariate logistics regression is used in those variables to further find out the independent risk factors relate to SLN status.Then R software was used to construct predictive models and nomograms to visualize the relationship between SLN status and independent risk factors.After the prediction model is obtained,the internal verification is carried out through 1000 times of Bootstrap resampling method.The H-L test and the calibration curve are used to evaluate the goodness of fit,and the receiver operating characteristic curve(ROC curve)area under the receiver(ROC curve)(AUC)is used to evaluate the prediction performance.Then substituting the verification group into the model for external verification,and also use AUC,H-L test,and calibration curve to evaluate the prediction power and fitness in the verification group.Results: 1.There is no statistical difference between the model group and the verification group in SLN results,clinical data,axillary ultrasound,and axillary MRI.2.The results of univariate logistics regression analysis show that:clinical data in the model group(tumor T staging),axillary ultrasound(lymph node cortex,lymph node hilum,lymph node blood flow pattern),axillary MRI(lymph node hilum,lymph node cortex,bilateral lymph nodes Symmetry or not),pathological results(pathological type,histological nuclear grade,lymphovascular invasion,perineural invasion)are closely related to SLN positive(P<0.05).All the statistically significant variables obtained from the above univariate logistics regression analysis are included in the multivariate logistics regression analysis to obtain: axillary ultrasound(lymph node hilum,lymph node blood flow pattern),axillary MRI(lymph node hilum,lymph node cortex),pathology results(histological nuclear grade,lymphovascular invasion,ER status)are important risk factors for predicting SLN metastasis.3.Construct a nomogram model for predicting the risk of SLN metastasis based on the feature variables selected by the above-mentioned multivariate logistics regression analysis.The internally verified AUC value of the model group in this study was 0.908,the 95% confidence interval was 0.868～0.949,and the H-L test showed a P value of 0.513(P>0.05),indicating that the model’s predictive ability and degree of fit were excellent.The data of the validation group(n=140)was substituted into the model,and the area under the ROC curve AUC was0.8736,and the H-L test showed a P value of 0.9733(P>0.05),indicating that the model has good predictive ability and good fit in the validation group.When the predicted risk of metastasis>0.812 is set to be positive for SLN,the diagnostic specificity of the model is 0.980,and the sensitivity is 0.706.Conclusion: 1.Logistic regression analysis showed that the preoperative pathological results(histological nuclear grade,lymphovascular invasion,ER status),axillary ultrasound(lymph node hilum,lymph node blood flow pattern),axillary MRI(lymph node hilum,lymph node cortex)and SLN The transfer status is closely related.2.The predictive model established by the above-mentioned related factors has a good predictive effect on the SLN status of patients with early breast cancer.3.The internal and external verification and goodness-of-fit test prove that the predictive ability and fit of the model are good. |
