| Background and purpose:The prevalence of breast cancer remains high and is a serious threat to women’s health.The number of cases of ductal carcinoma in situ(DCIS)is significantly higher than before,as a result of the malignant proliferation of ductal epithelial cells,that have not broken through the basement membrane and are confined to the ducts.DCIS is confined to the ductal-lobular system of the breast and does not invade the surrounding breast tissue,and has not yet infiltrated downwards to break through the basement membrane.However,in clinical practice,due to the variety of preoperative tissue sampling methods and the lack of precision in pathological diagnosis,the worst pathology of breast carcinoma is not accurately determined and therefore postoperative pathology escalation exists,and there is disagreement as to whether sentinel lymph node biopsy(SLNB)can be dispensed with during surgery for preoperatively diagnosed DCIS.Therefore,in order to draw reliable conclusions,to better guide clinical work and to provide patients with better access to care,this study attempts to identify risk factors for pathological underestimation of the initial diagnosis of DCIS and to develop a predictive model to screen appropriate patients with DCIS for SLNB in order to reduce the size of the patient’s surgical area involved and to reduce the corresponding number of patients.The present study sought to identify risk factors for underestimation of pathology in DCIS at initial diagnosis and to develop a predictive model to screen appropriate patients for exemption from SLNB in order to reduce the extent of surgical involvement and corresponding postoperative complications.Methods:Retrospective selection of all patients with DCIS who were initially diagnosed in Nanchang Third Hospital from January 2013 to December 2020 was conducted.Inclusion and exclusion criteria were set and relevant clinicopathological data were collected for patients who met the criteria.Univariate and multivariate logistic regression analyses were used to analyse the risk factors for pathological underestimation of DCIS,and dummy variables were reasonably set for multiple categorical variables.Regression analysis identified independent influencing factors for pathological underestimation.Based on this analysis result,a predictive model was constructed,and the predictive efficacy of the model was evaluated using subject performance characteristics(ROC)curve analysis to verify the clinical applicability of the model Results:A total of 826 DCIS patients who met the criteria were included.The results of univariate and multivariate logistic analysis suggested that lesion size,KI67,ER,HER2 status,histological grade,and diagnostic method were independent predictors of pathological underestimation(p<0.05).Based on the above research results,a prediction model was constructed: logit(P)=0.354 * lesion size+0.017 * KI67value+1.186 * ER-2.501 * diagnostic method(1)-1.575 * diagnostic method(2)-0.050 * HER state(1)-1.578 HER state(2)+1.160 * histological grade(1)+1.497 *histological grade(2)-2.418.The prediction model has a specificity of 73.8%,sensitivity of 79.2%,and accuracy of 76.2%.The area under the curve of the ROC curve is 0.856(95% CI: 0.831-0.881,P<0.001),indicating that the model has good predictive efficacy,confirming the superiority of our model in clinical application.To some extent,it can be used initially as a reference for exploring whether pathology is escalating Conclusion:The results of the regression analysis suggested that lesion size,KI67,ER,HER2 status,histological grading,and mode of diagnosis were independent influencing factors for pathological underestimation,while the ROC curve confirmed the good predictive efficacy of the prediction model constructed based on the analysis results of this study.Therefore,in clinical practice,to reduce postoperative-related complications and improve the quality of patient care,patients with preoperative diagnosis of ductal carcinoma in situ may be considered for intraoperative waiver of SLNB after considering the probability of pathological underestimation by this prediction model,provided that patients are adequately informed of the associated risks and that they give their informed consent. |
