| BackgroundEndometrial carcinoma(EC)is one of the common malignant tumors of female reproductive system,which is a serious threat to female health.High-grade EC including low differentiated endometrioid carcinoma(EEC)and non-endometrial carcinoma(NEEC).Compared with low-grade EC,high-grade EC is characterized by rapid progression,poor prognosis,and higher diagnostic heterogeneity.Systemic inflammatory response is one of the basic characteristics of malignant tumors.In patients with malignant tumors,the most direct and commonly used indicator of inflammatory response in the tumor microenvironment is the count of various inflammatory cells in peripheral blood,among which the count of neutrophils,lymphocytes and platelets in peripheral blood has been proved to be correlated with the prognosis of various types of malignant tumors.Neutrophil lymphocyte ratio(NLR)and Platelet lymphocyte ratio(PLR)are two hematological markers of systemic inflammatory response.Previous studies have suggested that NLR and PLR are closely related to the prognosis of patients with EC,but their correlation with the prognosis of patients with highgrade EC has not been clearly elaborated.This paper aims to explore the correlation between NLR and PLR and prognosis of high-grade EC patients,as well as the influence of other clinical and pathological factors on recurrence and death of high-grade EC patients,and to establish a prognostic model for high-grade EC patients based on the above characteristics.We expect to achieve risk quantification and risk stratification of patients’ recurrence and death through this model and realize visualization of the model and create a web version of the calculator,so as to facilitate clinicians to understand the prognosis of patients,optimize patient consultation,and provide reference for the development of follow-up plans.MethodsClinical data of 910 high-grade EC patients from 15 tertiary A medical centers in China were retrospectively analyzed.All patients were randomly assigned to either a modeling or validation group.1.Preliminary screening of variables and model construction:Disease-free survival(DFS)and Overall survival(OS)were used as study outcomes,respectively.Univariate Cox proportional hazard regression analysis(bilateral Log-rank test)was used for preliminary screening of all factors.Then,least absolute shrinkage and selection operator(LASSO)regression was used to screen clinical features.Subsequently,multi-factor Cox proportional risk regression was used to construct the risk assessment model of DFS and OS.Visualization of the model and scoring mechanism is achieved by nomograms.A prediction model web page calculator is made by Shiny app to simplify the use of the model and facilitate its promotion and application.2.Model evaluation and model verification:the area under the ROC(AUC)and concordance index were adopted as receiver operating characteristic curve(ROC).Concordance index(C-index)to evaluate the differentiation of the model.Calibration rate curves are used to evaluate the calibration degree of the model.The Decision Curve Analysis(DCA)was used to evaluate the benefit threshold probability of the model and explain the clinical net benefit of the model.Patients in the verification group were respectively used to verify the differentiation,calibration and clinical efficacy of the model.3.Model risk stratification:According to the scoring mechanism of the model,each patient was given a corresponding risk score for recurrence and death.X-tile was used to obtain the best truncation values of patients’ risk scores in the recurrence and death models,and patients were divided into high-risk group and low risk group according to the best truncation values.Kaplan-Meier(K-M)survival analysis was used to compare the difference between high and low risk groups.ResultsA total of 910 patients with high-grade EC were included in this study,including 724(79.5%)patients with poorly differentiated EEC,56(6.2%)patients with clear cell carcinoma(CCC),and 130(14.3%)patients with serous carcinoma(SC).All patients were randomly divided into the training cohort(639 cases)and the validation cohort(271 cases)at a ratio of 7:3.Chi-square test on the genetic characteristics of patients in the two groups showed no statistical significance.1.Univariate Cox proportional risk regression analysis and LASSO Cox multi-factor proportional risk regression analysis obtained:PLR,NLR,Types of Pathology,FIGO staging(International Federation of Gynecology and Obstetrics,Lymphovascular invasion(LVSI),lymphovascular invasion(LVSI),lymph node involvement and cervical involvement were independent factors affecting DFS.NLR,age,pathological type,lymph node involvement,depth of muscle invasion,cervical involvement,and FIGO stage were independent influencing factors for OS.A prediction model of mortality and recurrence risk was established based on the above characteristic variables,which was visualized by nomogram and simplified by web calculator.2.Model evaluation and verification results show that:In the DFS prediction model,the AUC values of ROC curve for 3,5 and 10 years in the modeling group were 0.85,0.85 and 0.86 respectively.The AUC values of ROC curve at 3,5 and 10 years in the verification group were 0.72,0.77 and 0.77,respectively.In the modeling group,the C-index of the patient DFS model was 0.83(95%CI 0.75-0.91).In the validation group,the C-index of the patient DFS model was 0.76(95%CI 0.62-0.91).In all patients,the C-index of the DFS model was 0.81(95%CI 0.74-0.88).In the histological subgroup,The Cindex of the DFS model of patients with poorly differentiated EEC group,serous carcinoma group and clear cell carcinoma group were 0.83(95%CI 0.72-0.95),0.80(95%CI 0.69-0.91)and 0.90(95%CI 0.70-0.90),respectively.In the modeling group,validation group,and histological subgroup,the results predicted by the model were in good agreement with the actual situation of patients,demonstrating the DFS prediction model had a good degree of differentiation.The calibration curve indicates the model has good calibration degree.DCA curve suggests this model has good clinical efficacy.In the OS prediction model,the AUC of the ROC curve at 3,5 and 10 years in the modeling group were 0.83,0.84 and 0.86,respectively.The AUC values of ROC curves at 3,5 and 10 years in the validation group were 0.72,0.81 and 0.84,respectively.In the modeling group,the C-index of the patient OS model was 0.85(95%CI 0.79-0.90).In the validation group,the Cindex of the patient OS model was 0.78(95%CI 0.68-0.89).In all patients,the C-index of the OS model was 0.80(0.74-0.86).In the histological subgroup,The C-index of OS model of patients in poorly differentiated EEC group,serous carcinoma group and clear cell carcinoma group were 0.85(95%CI 0.77-0.94),0.77(95%CI 0.64-0.91)and 0.82(95%CI 0.71-0.94),respectively.In the modeling group,validation group,and histological subgroup,the results predicted by the model were in good agreement with the actual situation of patients,proving the OS model had a good degree of differentiation.The calibration curve indicates the model has good calibration degree.DCA curve suggests this model has good clinical efficacy.3.K-M survival analysis of patients in high-risk and low-risk groups in the DFS and OS prediction models showed that:in the recurrence model,the survival difference between lowrisk and high-risk groups in the modeling group was statistically significant(P<0.001);In the verification group,the survival difference between low-risk group and high-risk group was also statistically significant(P=0.016).In the death model,the difference in survival between lowrisk group and high-risk group was statistically significant(P<0.001).In the verification group,the survival difference between low-risk group and high-risk group was also statistically significant(P=0.006).It is suggested that both DFS and OS models have good risk stratification and prediction ability.Conclusions1.NLR and PLR are risk factors affecting the prognosis of patients with high-grade EC.2.Combined with NLR/PLR and other clinicopathological features,a high-level EC prognostic risk prediction model was successfully constructed,which has a good calibration rate,discrimination and prediction accuracy,and can provide references for clinical diagnosis and treatment. |
PDF Full Text Request