Prediction Of Efficacy Of Preoperative Neoadjuvant Chemotherapy For Breast Cancer Based On MRI Features

Objective To analyze the factors and independent factors affecting the efficacy of neoadjuvant chemotherapy(NAC)and to construct efficacy prediction models based on multi-parametric magnetic resonance imaging(MRI)characteristics and clinical and pathological data of breast cancer patients before NAC.To explore the predictive efficacy and clinical application value of each model and provide a basis for personalized treatment of patients.Methods This study retrospectively collected the multi-parametric MRI features and clinical and pathological data of patients with invasive breast cancer diagnosed by pathological biopsy in General Hospital of Ningxia Medical University from October 2018 to March 2022,and evaluated the efficacy of NAC after surgical resection specimens.The response to NAC was evaluated by Miller-Payne grading system.G5 was defined as pathological complete response(pCR),and G1-G4 was defined as Non-pCR.According to the ratio of 7:3,the patients were divided into training set and validation set.Stepwise analysis of univariate and multivariate Logistic regression was used to screen the independent factors affecting the efficacy of NAC in the training set.the clinicopathological prediction model,imaging feature prediction model and combined prediction model were constructed respectively.Hosmer-Lemeshow test was used to test the goodness of fit of the models.Receiver operator characteristic curve(ROC)and calibration curve were used to evaluate the predictive performance of the three models.The area under the curve(AUC),sensitivity,specificity,accuracy,positive predictive value(PPV),negative predictive value(NPV)and F1 score was calculated.Delong test was used to compare the statistical difference of AUC of different prediction models.Decision curve analysis(DCA)was used to evaluate the clinical application value of each model,and a nomogram was drawn to visualize the optimal model.Results A total of 266 female breast cancer patients were included in this study,of which 95 cases achieved pCR and 171 cases did not achieve pCR after NAC.The overall pCR rate was 35.7%.1.Univariate and multivariate Logistic regression analysis showed that: menopausal status,NAC chemotherapy regimen,lymphocyte-to-monocyte ratio(LMR),molecular typing,estrogen receptor(ER),progesterone receptor(PR),human epidermal growth factor receptor2(HER2),Ki-67,and epidermal growth factor receptor(EGFR)had significant effects on pCR(P<0.05).LMR,ER,and HER2 were independent predictors of pCR(P<0.05).2.Univariate and multivariate Logistic regression analysis showed that: tumor diameter,margin,lobulation,spiculation,prepectoral edema and subcutaneous edema had statistically significant effects on pCR(P<0.05).Margin,lobulation and prepectoral edema were independent predictors of pCR(P<0.05).3.Construction and validation of clinicopathological prediction model,imaging feature prediction model and combined prediction model.The AUC values of the clinicopathological model in the training set and validation set were 0.788 and 0.721,the sensitivity was 0866 and0.714,the specificity was 0.583 and 0.706,and the accuracy was 73.26% and 70.89%,respectively.In the training set and validation set,the AUC values of the imaging feature prediction model were 0.682 and 0.743,the sensitivity were 0.463 and 0.857,the specificity were 0.833 and 0.549,and the accuracy were 70.05% and 70.89%,respectively.The AUC values of the combined prediction model in the training set and validation set were 0.842 and0.814,the sensitivity was 0.761 and 0.821,the specificity was 0.833 and 0.667,and the accuracy was 79.14% and 74.68%,respectively.The calibration curves of all models showed good prediction performance in the training set and validation set.Delong test and DCA were used to compare the benefits,and the prediction performance and clinical application value of the combined model were the best.Conclusion 1.Logistic regression analysis of clinicopathological and MRI imaging features showed that menopause,NAC chemotherapy regimen,LMR,molecular typing,ER,PR,HER2,Ki-67,EGFR,tumor diameter,margin,lobulation,spiculation,prethoracic edema and subcutaneous edema were related to pCR.Among them,LMR,ER,HER2,margin,lobulation and prepectoral edema were independent predictors.2.The clinicopathological prediction model,imaging feature prediction model and combined prediction model can be used to predict the efficacy of NAC before operation.The combined model combined with clinicopathological data and MRI imaging features has the best prediction efficiency and good clinical application value,which can provide effective reference for individualized treatment of breast cancer patients.