Construction And Validation Of A Predictive Model Of Pathological Complete Response Combined With MRI And Tumor Indicators For HER2-positive Breast Cancer After Neoadjuvant Targeted Therapy

Objective(s): Human Epidermal Growth Factor Receptor 2(HER2)positive breast cancer is a special subtype of breast cancer that is frequently characterized by poor biological manifestations including easier metastasis,worse prognosis and easier recurrence.However,the pathology Complete Response(p CR)rate of HER2 positive breast cancer was fairly high after standardized anti-HER2-targeted neoadjuvant therapy(NAT).p CR typically plays an significant role in directing the next stage of treatment.Therefore,by comparing general clinical indicators,imaging indicators,and hematological indicators of HER2-positive breast cancer patients between p CR group and non-p CR group,and then screen out independent predictors of p CR,this study aimes to establish a predictive model with a subsequent nomogram and then verified it so that we can use this model to predict the probability of reaching p CR in HER2-positive breast cancer patients at the stage of NAT and provide a reference for clinical treatment.Methods: Retrospectively and prospectively collected general clinical information,imaging data and hematological indicators in three different time nodes(baseline before NAT,after the first targeted therapy,and after the last NAT before surgery)of HER2-positive breast cancer patients receiving NAT in the Third Affiliated Hospital of Kunming Medical University(all baseline data below were labeled 0,the first neoadjuvant targeted therapy data were labeled 1,and the last neoadjuvant targeted therapy before surgery were labeled 2).Comparing the above data in the retrospective data between the p CR group and the non-p CR group,univariate logistic regression analysis combined with multivariate logistic regression analysis was performed to determine the independent predictors related to p CR.The prediction model was constructed and a nomogram was drawn.Internal validation was carried out through bootstrap resampling method,and the prediction performance of the model was evaluated according to calibration curve,DCA(Decision Curve Analysis)curve,ROC(Receiver Operator Characteristic curve)curve and AUC(Area Under the Curve).In addition,the ROC curve of the external validation set was drawn with prospective data to evaluate the prediction accuracy of the model in the external data.Results: 269 patients with HER2-positive breast cancer were retrospectively and 24 patients were prospectively included into this study.According to postoperative pathological conditions,the retrospective data were divided into p CR group and non-p CR group.18 factors related to p CR were screened out by univariable logistic regression analysis,and a total of 8 predictive factors affecting p CR were obtained by multivariable logistic regression analysis: PR(Progesterone Receptor)status(P=0.095),MRI2 ADC2(Apparent Diffusion Coefficient)(P=0.023),TIC2(Time Intensity Curve)(P < 0.001),ΔADC2(rate of ADC increase between ADC2 and ADC0)(P=0.023),ΔADC3(rate of ADC increase between ADC2 and ADC1)(P< 0.001),CEA-0(Carcinoembryonic Antigen)(P=0.078),CEA-1(P=0.039),CA153-2(Carbohydrate Antigen 153)(P=0.040).After constructing the multifactor analysis prediction model,we integrated the selected factors into a concise nomogram,which was verified internally through bootstrap method.The calibration curve showed that the model had good predictive ability for p CR,and the AUC value obtained according to ROC curve was 0.886(P < 0.001).The sensitivity was 0.822,and the specificity was 0.818,suggesting that this model has good prediction ability and high accuracy.According to the ROC curve of the external validation set,the calculated AUC value was 0.961(P< 0.001),the sensitivity was 1.000,and the specificity was 0.875,which further indicated that the diagnostic performance of the model was good.Conclusion(s): PR status,ADC2,TIC2,ΔADC2,ΔADC3,CEA0,CEA1,CA153-2 were independent predictors of attaining p CR after NAT in patients with HER2-positive breast cancer.The nomogram constructed based on the results of multivariable logistic regression in this study can be applied to predict the post-NAT p CR rate of HER2-positive breast cancer patients,and the internal and external verification proves that the model has high prediction accuracy and good diagnostic prediction efficacy,which can provide reference and inspire new ideas for guiding the clinical precise individual chemotherapy treatment.