| Objective To analyze the clinical characteristics of platinum resistance in epithelial ovarian cancer and build a prediction model based on clinical characteristics;Screening platinum-resistant genes in epithelial ovarian cancer and constructing a prediction model based on the immunohistochemical(IHC)score of platinum-resistant genes;The platinum resistance model of epithelial ovarian cancer was optimized by combining the clinical characteristics and the platinum resistance gene IHC score to provide a basis for the prediction of platinum resistance in clinical epithelial ovarian cancer.Methods 1.From January 31,2019 to December 31,2020,124 patients with epithelial ovarian cancer(EOC)who underwent initial treatment(satisfactory tumor cell reduction combined with chemotherapy)in the General Hospital of Ningxia medical University were collected.The patients were divided into platinum-sensitive group and platinum-resistant group according to their response to chemotherapy during initial treatment.General information such as age,menopausal status,ascites,tumor unilateral or bilateral,tumor size.Pathological characteristics such as clinical stage(FIGO stage),pathological type,histological grade,lymph node metastasis,vascular tumor thrombus,ascites cancer cells,therapeutic regimen such as neoadjuvant chemotherapy,surgical method,postoperative residual lesions and chemotherapy cycle.Tumor markers include: CA125,HE4,CA199,CEA,AFP;Inflammatory indicators: Neutrophil-lymphocyte-count ratio(NLR),Platelet-lymphocyte-count ratio(PLR),Monocyte-lymphocyte-count ratio(MLR),Albumin(ALB),Albumin to globulin ratio(A/G),univariate analysis of clinical characteristics,The prediction factors were screened by using the lasso regression(LASSO)analysis.The prediction model is established through the logistic regression analysis,and the nomogram function is used to draw the nomogram.Calculate the consistency index(C-index),draw the calibration curve,the receiver operating characteristic(ROC)curve,and the decision curve analysis(DCA)to analyze and evaluate the performance of the model,and internally verify the prediction value of the model through the sampling method.2.Through the GEO database GSE45553,GSE28739 data set screened the significantly different genes based on the criteria of | log FC | > 1.0,p < 0.05,analyzed the interaction(PPI)network between gene coding proteins through the STRING online database,and further clustered the functional module through the Cytascape software MCODE plug-in,and selected the genes in the sub-network with the highest score as the Hub gene.In addition,nine classical platinum resistance genes CTR1,P-gp,ABCC1,BCRP,LRP,BCL2,P53,Survivin,ERCC1 were screened by literature method.There were 70 tumor tissue samples of EOC patients,including 43 platinum-sensitive patients and 27 platinum-resistant patients.The platinum-resistant gene was verified by IHC,and the results of the platinum-resistant gene IHC test were automatically scored by the IHC Profiler plug-in of Image J software.The prediction model based on platinum resistance gene IHC score was constructed and evaluated.3.The 70 patients with EOC collected from January 31,2019 to December 31,2020 were used as the training set for model optimization.The 7 clinical factors included in the prediction model of EOC based on clinical characteristics and the 6 platinum resistance genes included in the prediction model of epithelial ovarian cancer based on the IHC score of platinum resistance genes were screened by LASSO regression.The optimized prediction model is constructed and evaluated..Collect the patients with EOC who visited the General Hospital of Ningxia Medical University from January 31,2021 to June 30,2021 as the test set,predict the platinum resistance risk of the patients in the test set according to the optimization model,predict the platinum resistance and platinum sensitivity according to the cut-off value obtained from the training set,evaluate the authenticity of the optimization model by drawing the ROC curve,calculating the AUC,sensitivity,specificity and likelihood ratio of the prediction model,and calculate the coincidence rate The consistency test(Kappa test)evaluates the reliability of the optimization model.Results1.Establishment and evaluation of platinum resistance prediction model for EOC based on clinical characteristics:(1)Univariate analysis of the clinical characteristics of EOC showed that ascites(p=0.027),tumor size(p=0.033),vascular tumor thrombus(p=0.001),lymph node metastasis(p=0.038),ascites cancer cells(p<0.001),neoadjuvant chemotherapy(p<0.001),and residual lesions(p=0.017)were associated with platinum resistance in epithelial ovarian cancer.(2)LASSO regression analysis screened seven predictors of clinical characteristics of platinum resistance in epithelial ovarian cancer,including vascular tumor thrombus,ascites cancer cells,tumor size,neoadjuvant chemotherapy,residual lesions after surgery,HE4 and CA199,with OR values(95% CI)of 3.489(1.375-9.307),3.171(1.085-9.793),0.786(0.293-2.152),2.385(0.856-6.764),3.549(0.756-19.378),1.002(1.000-1.003),0.938(0.800-1.015),respectively.(3)The regression coefficients of vascular metastasis,ascitic cancer cells,tumor size,neoadjuvant chemotherapy,residual lesions after surgery,HE4 and CA199 were 1.250,1.153,-0.241,0.869,1.267,0.002 and-0.064,respectively,using Logistic regression analysis.The prediction model of platinum resistance of EOC based on clinical characteristics was constructed and visualized by nomogram.The prediction model C-index is 0.850,and the internal sampling inspection C-index is 0.800.The calibration curve of the model shows that the prediction model is highly consistent,and the ROC curve shows that the model has good discrimination ability.The clinical decision curve shows that the model has the maximum clinical income within the threshold of 1%-100%.2.Construction and evaluation of platinum resistance risk prediction model for EOC based on platinum resistance gene IHC score:(1)The GEO data sets GSE45553 and GSE28739 screened 56 platinum-resistant and significantly different genes in EOC,of which47 were up-regulated and 9 were down-regulated.Nine Hub genes were obtained through the cluster analysis of Cytascape software MCODE plug-in: SOCS3,CEBPB,IL1 B,CXCL1,CXCL2,IL6,DUSP1,NFKBIZ,LIF,which were verified by IHC,including CXCL1,CXCL2,IL6,DUSP1 NFKBIZ was highly expressed in platinum-resistant epithelial ovarian cancer tissues,and the difference was statistically significant(p<0.05).(2)The classical platinum-resistant genes were screened by literature method and the results of IHC showed that CTR1 was highly expressed in platinum-sensitive tissues,and the expression of P-gp,ABCC1,BCRP,LRP,BCL2,P53,survivin,ERCC1 in platinum-resistant tissues was significantly higher than that in platinum-sensitive tissues(p<0.05).(3)The IHC score of platinum-resistant genes CXCL1,CXCL2,IL6,ABCC1,LRP,and BCL2 in EOC screened by LASSO regression were included in the prediction model.The OR(95% CI)of the IHC scores of IL6,CXCL1,CXCL2,ABCC1,LRP,and BCL2 were 4.539(1.184-21.384),4.879(0.734-38.827),5.293(1.047-32.286),3.429(0.535-24.940),1.722(0.329-9.824),2.028(0.428-22.353),respectively.(4)Through logistic regression analysis,the regression coefficients of the IHC scores of platinum-resistant genes IL6,CXCL1,CXCL2,ABCC1,LRP,and BCL2 were 1.513,1.585,1.666,1.232,0.544,and 1.074,respectively.The risk prediction model of platinum-resistant EOC based on the IHC scores of platinum-resistant genes was constructed and visualized in the form of nomogram.The C-index of the prediction model is 0.939,and the internal sampling test results show that the C-index is 0.916.The calibration curve of the model shows that the prediction model is highly consistent.The ROC curve shows that the discrimination ability of the model is better than that of the model based on clinical characteristics.The clinical decision curve indicates that there is the maximum clinical net income within the threshold of 4%-100%.3.Optimization and validation of platinum resistance prediction model for EOC:(1)The optimal prediction model of platinum resistance in EOC was constructed by combining clinical characteristics and the IHC score of platinum resistance gene.The LASSO regression analysis included 11 predictive factors including the IHC score of platinum resistance gene CXCL1,CXCL2,IL6,ABCC1,LRP,BCL2,and vascular tumor thrombus,ascites cancer cells,tumor size,neoadjuvant chemotherapy,HE4.(2)Through logistic regression analysis,the regression coefficient of vascular metastasis,ascites with cancer,neoadjuvant chemotherapy,tumor size,serum HE4 were 2.359,1.157,1.691,-1.337,0.001,respectively.The regression coefficient of IHC scores of IL6,CXCL1,CXCL2,ABCC1,LRP,BCL2 were1.582,2.834,2.661,2.149 0.615 、 0.215.The optimization prediction model equation is constructed and visualized through the form of nomogram.The C-index of the optimized prediction model is 0.972,and the cut-off value of the total points of the nomogram calculated according to the AUC curve is 163.0.At this time,the sensitivity of the prediction model is95.35%,the specificity is 88.89%,and the positive likelihood ratio is 8.581.(3)The result of external validation of the optimized prediction model through the test set of patient data shows that the coincidence rate is 84%.The consistency test Kappa is 0.677,indicating strong consistency.The ROC curve showed that the AUC of the prognosis model was 0.930(95% CI0.833-1.000),the sensitivity of the prediction model was 95.35%,the specificity was 88.89%,and the positive likelihood ratio was 5.96.Conclusion 1.The prediction model of EOC based on clinical characteristics includes seven factors,including vascular metastasis,ascites cancer cells,tumor size,neoadjuvant chemotherapy,residual lesions after surgery,serum HE4 and CA199.The model has good comprehensive application performance.2.The platinum resistance risk prediction model of EOC based on the IHC score of platinum resistance gene constructed by using the GEO database and the platinum resistance gene screened by the literature method includes the IHC scores of CXCL1,CXCL2,IL6,ABCC1,LRP and BCL2.This model has higher clinical application value than the prediction model based on clinical characteristics.3.The vascular metastasis,ascites cancer cells,tumor size,neoadjuvant chemotherapy,HE4 and IHC scores of CXCL1,CXCL2,IL6,ABCC1,LRP,BCL2 were included in the optimized platinum resistance prediction model of EOC.Compared with the risk prediction model of platinum resistance in EOC based on clinical characteristics and the IHC score of platinum resistance gene,the optimized prediction model of platinum resistance in EOC has the highest clinical application value,and the external validation results also indicate that the model has good universality. |
PDF Full Text Request