Clinical Study Of Tumor-Infiltrating Lymphocytes In Breast Cancer After Neoadjuvant Therapy

Neoadjuvant therapy for breast cancer refers to the method of systemic treatment before local treatment of breast cancer patients,and is currently an important means of breast cancer treatment.Visual quantitative assessment of tumor-infiltrating lymphocytes(TILs)after neoadjuvant therapy for breast cancer lacks accuracy and repeatability.Studies have shown that TILs and TILs subtypes have extremely high prognostic value in breast cancer after neoadjuvant therapy.Therefore,we compared the difference and consistency of TILs after neoadjuvant therapy by microscopic visual assessment(VA)and artificial intelligence(AI)to improve the accuracy of TILs interpretation,and then studied the prognostic value of TILs and TILs subtypes in breast cancer patients after neoadjuvant therapy.Part one Repeatable study of tumor-infiltrating lymphocytes in breast cancer after neoadjuvant therapy based on artificial intelligenceObjective: This study compared the difference and repeatability of VA and AI assisted interpretation of TILs of breast cancer after neoadjuvant therapy under microscope,and explored the clinical applicability of AI combined with pathologists to interpret TILs of breast cancer after neoadjuvant therapy.Methods: This study included 50 patients with invasive breast cancer who were diagnosed in the fourth hospital of Hebei Medical University from2014 to 2015 and underwent surgical resection after neoadjuvant therapy.All patients did not obtain pathological complete response(p CR).In this study,the AI algorithm used cell detection and region segmentation techniques to quantitatively interpret the TILs in the H&E digital image.Two pathologists with more than 15 years of working experience who did not participate in this study used the common interpretation results of LCA immunohistochemical sections by multi-eye microscope as the gold standard of this study.Nine pathologists of different levels used evaluate TILs of breast cancer after neoadjuvant therapy by VA and AI-assisted.In this study,SPSS 26.0 and Graph Pad Prism 8.0.1 were used for statistical analysis.Friedman M and Bonferroni correction are used for difference analysis.Intraclass correlation coefficient(ICC)and Bland-Altman scatter plot were used for consistency analysis.Results:1.Difference analysisNine pathologists in the VA group and the AI-assisted group found that the TILs of 50 cases of breast cancer after neoadjuvant therapy did not follow the normal distribution after normality test(P < 0.05).The Friedman M test further showed that the interpretation results of the 9 pathologists in the VA group were significantly different(P < 0.001),while the interpretation results of the 9 pathologists in the AI-assisted group were no significant difference(P= 0.416)for the TILs interpretation of breast cancer after neoadjuvant therapy.At the same time,the TILs interpretation results of pathologists were compared by Bonferroni correction.There was no significant difference between senior pathologists in the VA group in the TILs interpretation of breast cancer after neoadjuvant therapy(P > 0.05).The TILs interpretation results were significantly differences between senior pathologists and intermediate and junior pathologists(P < 0.05),and there were also significant differences in the TILs interpretation results of between intermediate and junior pathologists(P < 0.05).2.Consistency analysis1)Through the consistency test,the ICC of TILs interpretation between senior pathologists in VA group was 0.842(95% CI 0.762-0.901),with good consistency;the ICC between intermediate pathologists was 0.735(95% CI0.617-0.829),and that between junior pathologists was 0.653(95% CI0.513-0.771),with poor consistency.The results showed that the consistency of TILs interpretation by senior pathologists was higher than that by intermediate and junior pathologists in VA group.2)The ICC of TILs interpretation for breast cancer after neoadjuvant therapy between AI and the gold standard is 0.977(95% CI 0.961-0.987),which is in excellent agreement and higher than the agreement between all pathologists and the gold standard.3)The ICCs of TILs interpretation for breast cancer after neoadjuvant therapy between all pathologists in AI-assisted group and the gold standard were higher than 0.9,and they all had excellent consistency.Part two Constructing a prognostic model of breast cancer after neoadjuvant therapy based on tumor-infiltrating lymphocytes and its subtypesObjective: In this study,AI was used to interpret the TILs of breast cancer after neoadjuvant therapy,and then to evaluate the prognostic value of TILs,TILs subtypes and clinicopathological indicators.The nomogram prognostic model was constructed by integrating factors with independent prognostic value to predict the overall survival(OS)and disease-free survival(DFS)of breast cancer after neoadjuvant therapy.Methods: A retrospective study of patients with invasive breast cancer diagnosed by preoperative core needle biopsy from January 2013 to December2015 in the Fourth Hospital of Hebei Medical University and with surgical resection after neoadjuvant therapy were screened out and the clinicopathology data was improved.A total of 209 patients were eligible for enrollment,and all patients received 6-8 cycles of neoadjuvant chemotherapy and targeted therapy.The patients were randomly divided into training set and validation set,and Cox proportional hazard regression was used for survival analysis,and nomograms were drawn to predict the prognosis of breast cancer after neoadjuvant therapy.The calibration chart examines the consistency of the nomograms,and the C-index and Time-ROC curve examine the degree of discrimination of the nomograms.Results:1.Among the TILs subtypes,CD45 RO is the most cell type,with an average of 25%,followed by CD3 18%,CD8 11%,CD4 9% and Fox P3 5%.There were no significant differences in TILs and TILs subtypes in each molecular type.2.TILs and its subtypes were positively correlated(P < 0.01),and negatively correlated with histological grade,lymph node status,clinical stage and lymphatic vascular infiltration(P < 0.01).3.TILs and its subtypes are negatively correlated with residual cancer burden(RCB)and p CR,and the correlation is statistically significant(P <0.01).4.In the training set,TILs infiltration degree,expression of CD4 and Fox P3 after neoadjuvant therapy,metastasis lymph node ratio(LNR),clinical stage and HER2 expression before neoadjuvant therapy were independent influencing factors of OS(P < 0.05).TILs infiltration degree,expression of CD4 and Fox P3 after neoadjuvant therapy and clinical stage were independent influencing factors of DFS(P < 0.05).5.Construction and validation of prognosis model1)Combined factors with independent prognostic value to construct a nomogram prognostic model to predict OS rate and DFS rate respectively.2)Consistency test: In the training set and validation set,compared with the ideal model,the calibration chart of nomogram for predicting OS rate and DFS rate shows good consistency.3)Discrimination test: In training set,the C-index for predicting the OS rate is 0.909,and the area under the Time-ROC curve is higher than 0.9;the C-index for predicting the DFS rate is 0.878,and the area under the Time-ROC curve is also higher than 0.9.In validation set,the C-index for predicting the OS rate is 0.886,and the area under the Time-ROC curve is higher than 0.8;the C-index for predicting the DFS rate is 0.869,and the area under the Time-ROC curve is higher than 0.75,indicating that the nomograms for predicting OS rate and DFS rate have good accuracy.4)Decision curve analysis: Compared with the single significant predictor model,the prediction of OS and DFS in breast cancer after neoadjuvant therapy based on the nomogram can obtain a greater net benefit.Conclusions:1.AI significantly improves the consistency and repeatability of the interpretation results of TILs by pathologists.It can be seen that AI-assisted pathologists is a good way to improve the consistency and repeatability of the TILs interpretation results of breast cancer.2.The higher the TILs infiltration degree and the expression of CD3,CD4,CD45 RO,CD8 and Fox P3 positive cells,the higher the p CR rate after neoadjuvant treatment of breast cancer and the lower the RCB class.3.The higher the TILs infiltration degree,the expression levels of CD4 and Fox P3 in breast cancer after neoadjuvant therapy and HER2 positive expression before neoadjuvant therapy,the higher the OS rate and the better the prognosis of patients;while the higher the clinical stage and the LNR,the lower the OS rate and the worse the prognosis of patients.4.The higher the TILs infiltration degree and the expression levels of CD4 and Fox P3 in breast cancer after neoadjuvant therapy,the higher the DFS rate and the better the prognosis of patients;while the higher the clinical stage,the lower the DFS rate and the worse the prognosis of patients.5.The nomograms for predicting OS rate and DFS rate have good discrimination and consistency,which can provide predictive value for the prognosis of breast cancer after neoadjuvant therapy.