Building A Prediction Model And Nomogram For Efficacy Of Neoadjuvant Chemotherapy

Objective: Neoadjuvant chemotherapy(NAC)for breast cancer is the standard treatment regimens for locally advanced breast cancer and inflammatory breast cancer,and its application in early breast cancer has become more and more common.NAC can reduce the stage of tumor,provide a support for completely excising of tumor,increase the breast and axillary conservation rates,and determine the sensitivity of the tumors to chemotherapeutic drugs,and lay the foundation for optimizing and selecting the most scientific treatment scheme.However,there are still a few patients who are not sensitive to NAC,and even NAC not only fails to achieve the expected effect,but also forms irreversible toxic and side effects,which may cause the patients to lose the best window of surgical intervention and bring poor prognosis.Therefore,accurately predicting the efficacy of NAC contributes to screening the NAC candidates and providing a reference for the selection of the treatment.This study combined the clinical factors with pathological factors to deeply investigate the factors related to the efficacy of NAC,and built a predication model to improve the effectiveness and accuracy of predicting the efficacy of NAC,thus providing a reference for the selection of the clinical diagnosis and treatment.Method: The clinicopathological data of 402 breast cancer patients who received NAC in Breast Center of the Affiliated Hospital of Qingdao University from January 1,2020 to December 31,2021 were retrospectively analyzed.They were randomly divided into training set and test set according to the ratio of 7:3.According to the MP grading system post NAC,the patients were divided into effective group and ineffective group.The prediction variables were screened by two methods,the first one was to use χ~2 test for single-factor analysis,the statistically significant and possibly clinically significant indicators were included in multi-factor logistic regression.The second one is LASSO regression,and the variables screened by LASSO regression were used to build the prediction model,which is shown in the form of nomogram.The receiver operating characteristic curve(ROC)was used to evaluate the performance of the model,the Hosmer-Lemeshow test and calibration curve were used to evaluate the accuracy of the model.The decision curve analysis(DCA)was used to evaluate the clinical application value of the model.At last,the model was internally verified in the test set.Results:1.A total of 402 patients undergoing NAC were included in this study,there were 301 cases in the effective group,101 cases in the ineffective group,282 cases in the training set,and 120 cases in the test set.The single-factor analysis showed that the age,menstrual status,BMI,tumor size,chemotherapy regimen,ER,PR,HER-2,Ki-67,molecular classification and RECIST 1.1 evaluation after four cycles were associated with the efficacy of NAC(all P < 0.05),while the lymph node status,chemotherapy cycle and histological grade were not associated with the efficacy of NAC(all P > 0.05).2.The multi-factor regression showed that,six independent variables were included in the multi-factor logistic regression-based prediction model,including tumor size,chemotherapy regimen,histological grade,Ki-67,molecular classification and RECIST1.1 evaluation after four cycles.Eight variables were included in the LASSO regression-based prediction model,including tumor size,chemotherapy regimen,histological grade,ER,HER-2,Ki-67,molecular classification and RECIST 1.1evaluation after four cycles.The area under the curve(AUC)of two models were 0.812 and 0.875 respectively,and the net reclassification improvement(NRI)and integrated discrimination improvement(IDI)were 0.141 and 0.117 respectively.3.Based on the results of LASSO analysis,the statistically significant variables were used to build the model and nomogram.The calibration curve and Hosmer-Lemeshow test showed a better goodness of fit of the model,and the DCA curve showed that the model had a clinical application value.The AUC of the test set was 0.792 in internal verification.4.The Kappa consistency test between RECIST 1.1 evaluation after four cycles and pathological assessment showed Kappa = 0.377,P < 0.001.The AUC of RECIST 1.1evaluation after four cycles alone for predicting the efficacy of NAC was 0.694,and its difference from that of the prediction model was statistically significant as shown by the Delong test(z =-7.22,P < 0.001).In the single-variable prediction model of RECIST 1.1after four cycles,the NRI and IDI were-0.216 and-0.236 respectively.Conclusion1.The prediction model built with a total of 8 predictors including tumor size,chemotherapy regimen,histological grade,ER,HER-2,Ki-67,molecular classification and RECIST 1.1 evaluation after four cycles,has a good predictive ability for the efficacy of NAC.2.RECIST 1.1 after four cycles showed a poor Kappa consistency with the pathological assessment,and the single-factor prediction model showed a general predictive ability...