Prediction Of Axillary Lymph Node Metastasis In Breast Cancer Based On Mammographic Mass Characteristics

Part1 Prediction of Axillary Sentinel Lymph Node Metastasis in Breast Cancer Based on Mammographic Mass CharacteristicsObjective:Combined with the characteristics of mammography(MG)mass,immunohistochemistry and general conditions of patients,the risk factors of axillary sentinel node(SLN)metastasis in patients with clinically early-stage breast cancer(c T1-T2)were analyzed to establish a clinical prediction model for SLN metastasis,and the predictive efficacy and clinical practicability of the model were assessed.Materials and methods:Mammography,immunohistochemistry and general conditions of patients with clinically early-stage breast cancer(c T1-T2)who underwent sentinel lymph node biopsy(SLNB)in our hospital from January 2018 to March 2021were collected.Clinical palpation,mammography and ultrasonography showed negative axilla in all patients.Collected variables included the following:mammary gland(fatty,fibroglandular,nonuniform dense,extremely dense),mass density(high density,equal density),mass shape(round or oval,irregular),mass border(circumscribed,obscured,indistinct),spiculation,microlobulation,calcification,mass size,ER,PR,HER-2,Ki-67,patient age,and menstrual status(premenopausal,postmenopausal).A total of 649patients were included in the study according to the screening criteria and divided into training group(n=487)and validation group(n=162)according to the ratio of 3:1.Univariate logistic regression analysis was performed on all independent variables included in the study in the training group to screen independent variables that were statistically associated with SLN metastasis and included them in multivariate logistic regression analysis,to investigate independent risk factors with SLN metastasis according to the results of multivariate analysis,and to establish a SLN metastasis prediction model.The area under the ROC curve,calibration curve,and decision curve were used to assess the discrimination,calibration,and clinical application value of the model,respectively,and external validation was performed with validation group data.Results:1.Univariate analysis showed that there were significant differences in circumscribed border(P=0.002),indistinct border(P<0.001),density(P=0.037),shape(P<0.001),spiculation(P<0.001),calcification(P<0.001),mass size(P<0.001),PR(P=0.018)and menstrual status(P=0.065)between the two groups,which were correlated with axillary SLN metastasis.2.Multivariate analysis suggested that there was a statistical association between mass border,spiculation,calcification,mass size and menstrual status and SLN metastasis.Compared with obscured border,indistinct border will increase the risk of SLN metastasis(OR=4.22,95%CI:2.17–8.24,P<0.001),while circumscribed border has a lower risk(OR=0.32,95%CI:0.12–0.85,P=0.023);Compared with non-spiculated masses,spiculated masses will increase the risk of SLN metastasis(OR=2.36,95%CI:1.38-4.03,P=0.002);compared with non-calcified masses,calcified masses will increase the risk of SLN(OR=2.23,95%CI:1.41-3.51,P=0.001);for each unit increase in mass size,the probability of SLN metastasis will also increase(OR=3.60,95%CI:2.50-5.19,P<0.001);relative to postmenopausal patients,premenopausal patients have a higher risk of SLN metastasis(OR=1.83,95%CI:1.10-3.05,P=0.021).A nomogram prediction model was established according to the above indicators,and the area under the ROC curve of the obtained model was 0.835(95%CI:0.799-0.871).The calibration plot showed that the fitting curve was close to the ideal curve and ran in accordance with it.X~2=12.191,P=0.143 of the Hosmer-Lemeshow fit goodness test,and the model fit was good.The decision curve showed threshold probabilities of net benefit for patients ranging from2%to 85%and 90%to 98%.The area under the ROC curve of the external validation model was 0.828(95%CI:0.766-0.890),and the calibration curve of the validation group showed good model fit.Conclusion:1.Mass density,shape,border,spiculation,calcification,mass size,PR and menstrual status are correlated with SLN metastasis in early breast cancer,and mass border,spiculation,calcification,mass size and menstrual status are independent risk factors for SLN metastasis.2.The model has high discrimination and calibration and wide clinical applicability,which can provide some reference for clinicians to make treatment decisions.Part2 Prediction of Axillary Non-Sentinel Lymph Node Metastasis in Breast Cancer Based on The Characteristics of Mammography Mass and Sentinel Lymph NodeObjective:Based on mammographic mass characteristics,immunohistochemistry,SLN characteristics and the general condition of patients with clinically early-stage breast cancer(c T1-T2),the factors associated with axillary non-sentinel lymph node(NSLN)metastasis were analyzed,and a single-index model labeled with mammographic mass characteristics,a single-index model labeled with SLN characteristics,and a combined model integrating mammographic mass characteristics and SLN characteristics were established,respectively,to assess and compare the predictive efficacy and clinical practicability of the three models.Materials and methods:The clinical data of patients with early breast cancer who underwent SLNB+ALND in our hospital from January 2018 to March 2021 were collected.Clinical palpation,mammography and ultrasonography showed negative axilla in all patients.Collected variables included the following:mammary gland(fatty,fibroglandular,nonuniform dense,extremely dense),mass density(high density,equal density),mass shape(round or oval,irregular),mass border(circumscribed,obscured,indistinct),spiculation,microlobulation,calcification,mass size,ER,PR,HER-2,Ki-67,patient age,and menstrual status(premenopausal,postmenopausal),ER,PR,HER-2,Ki-67,lymph node information(SLN metastatic status,SLN macrometastasis,number of SLN positive,number of SLN negative,SLN positive rate,NSLN metastatic status).A total of 271 patients were included in the study according to the screening criteria and divided into training group and validation group according to 3:1,including 199 patients in the training group and 72 patients in the validation group.Pearson’s chi-square test,Fisher’s exact test and independent samples t-test were used for univariate analysis of all independent variables included in the study in the modeling group to screen for factors associated with NSLN metastasis.To avoid the possible linear correlation between independent variables,multivariate regression analysis was then performed using cross-validated Lasso-logistic to screen independent risk factors for NSLN metastasis.According to the results of multivariate regression,the single-index model with mammographic mass characteristics as the label,the single-index model with SLN characteristics as the label,and the combined model integrating mammographic mass characteristics and SLN characteristics were established,respectively,and the discrimination,stability,calibration,and clinical practicability of the three models were assessed and compared using ROC curve,Akaike information criterion(AIC),calibration curve,and decision curve,respectively,and the degree of model power improvement was assessed with the comprehensive discriminant index(IDI),and external validation was performed with the validation data.Results:1.Univariate analysis suggested that mass shape(P=0.005),mass border(P<0.001),spiculation(P=0.001),mass size(P<0.001),calcification(P=0.037),SLN metastatic status(P=0.001),number of SLN positives(P<0.001),and SLN positive rate(P=0.001)were statistically different between the two groups and correlated with NSLN metastasis.2.Multivariate regression suggested that mass border,mass size and SLN macrometastasis,and the number of positive SLNs were independent risk factors for NSLN metastasis.Compared with obscured border,indistinct border increased the risk of NSLN metastasis(OR=3.15,95%CI:1.01-9.85,P=0.049);for each unit increase in mass size,the probability of NSLN metastasis will also increase(OR=3.14,95%CI:1.76-5.62,P<0.001);those with SLN macrometastasis increased the risk of NSLN metastasis relative to those without SLN macrometastasis(OR=7.85,95%CI:3.05-20.21,P<0.001);and those with≥3 SLN positives increased the risk of NSLN metastasis relative to<3 SLN positives(OR=10.77,95%CI:2.65-43.87,P=0.001).A single-index model(MG model)labeled with mammographic mass characteristics,a model labeled with SLN characteristics(SLN model),and a combined model were established,respectively,and the analysis results suggested that the areas under the ROC curves of the combined model,MG model,and SLN model in the training group were 0.864(95%CI:0.815-0.913),0.772(95%CI:0.710-0.836),and 0.792(95%CI:0.740-0.844),respectively,and those in the validation group were 0.878(95%CI:0.802-0.954),0.827(95%CI:0.739-0.916),and 0.774(95%CI:0.686-0.862),respectively.Using IDI to express the comprehensive discriminant ability of the model,the comprehensive discriminant ability of the combined model in the training group was increased by 17%and 10%,respectively,compared with the MG model and SLN model,and 10.7%and17.3%in the validation group,respectively,which were positive improvements,and the results were statistically significant(P<0.001).The stability of the three models was tested by AIC,and the smaller the AIC,the better the stability of the model.The AIC of the combined model,MG model and SLN model in the training group were 185.36,222.47 and 207.65,respectively,while those in the validation group were 78.01,84.00and 91.92,respectively.The stability of the combined model in both groups was the best.The calibration curve and decision curve showed that the three models had good calibration fit and clinical practicability,and the calibration and overall net benefit of the combined model of the training group and the validation group were better than those of the other two single-index models.Conclusion:1.Mass shape,border,spiculation,calcification,mass size,SLN metastatic status,SLN macrometastasis,number of SLN positive,and SLN positive rate were significant factors for NSLN metastasis.Among them,mass border,mass size,SLN macrometastasis,and number of positive SLN were independent risk factors for NSLN metastasis.2.The three models established according to multivariate analysis had good predictive ability,and the combined model showed superior predictive performance and optimal stability than the other two single-index models.