Development And Validation Of A Stratified Analysis And Predictive Model Of Risk Factors For Small Tumour Lymph Node Metastasis In Primary Breast Cancer

OBJECTIVE: The aim of this study was to conduct a retrospective analysis to develop a clinical predictive model to predict risk factors associated with lymph node metastasis in patients with primary breast cancer with small tumours and to explore the implications of its clinical use.METHODS: The medical record data of patients diagnosed with stage T1 breast cancer from January 2011 to December 2020 at the Cancer Hospital of Xinjiang Medical University(N=2370)were retrieved,those who did not meet the criteria(N=1082)were excluded,and those cases meeting the inclusion and exclusion criteria(N=1288)were further divided into the lymph node metastasis group(N=490)and the no lymph node metastasis group(N=798)to collect General patient information,clinicopathological data and laboratory data were collected.The diagnosis of anterior and axillary lymph node metastases was mainly based on postoperative paraffin pathological examination.The data set of the lymph node metastasis group was randomly sampled into two groups(one for modelling and one for validation)in the ratio of 7:3 by using R software(version 4.1.0)"caret",setting a random seed,and using the create Data Partition function,while the group without lymph node metastasis was randomly sampled into two groups(one for modelling and one for validation)in the ratio of 7:3.The data set of the lymph node metastasis group was randomly sampled into two groups(control group and validation group two)in the ratio of 7:3,while the group without lymph node metastasis was randomly sampled into two groups(control group and validation group two).The data were analysed using SPSS software(version 26.0)in the modeling group to explore the adverse effects of lymph node metastasis and to develop a predictive model.The validation set was further investigated using logistic regression to explore the risk factors associated with lymph node metastasis.The validation set was used to test the predictive model and the multi-factor logistic regression model separately,and the two established models were compared and tested using the subject operating characteristic curve(ROC).RESULTS: A total of 1288 cases met the requirements of this study,of which 490 were positive for lymph node metastasis.Analysis comparing the lymph node metastasis group with the non-metastasis group showed that variables such as maximum tumour diameter size,vascular cancer thrombus,nerve invasion and Ki67 were significantly associated with lymph node metastasis and were statistically different.Logistic regression analysis showed an AUC of 0.695 for the modelling group and 0.642 for the validation group,with a similar level of model prediction.Conclusion: There are many adverse factors for lymph node metastasis,and the important factors screened in this study were tumour size,molecular staging,choroidal carcinomatosis invasion,tumour grade and BMI.In the multifactorial logistic regression analysis,tumour size,choroidal carcinomatosis invasion,nerve invasion and Ki67 positivity were the independent risk factors for ALNM.In addition,the line graphs developed in this study showed good predictive ability to accurately assess the risk factors of patients with lymph node metastasis,which is useful for individualising the management of patients with stage T1 breast cancer and minimising overtreatment.