Construction And Evaluation Of A Prognostic Prediction Model For Adult Patients With Low-grade Glioma

Background & objective:Glioma is the most common intracranial malignant tumor,accounting for about33% of intracranial tumors.According to the World Health Organization((World Health Organization,WHO)brain glioma grading standards,WHO I grade and WHO II grade are low grade glioma(Low-grade glioma,LGG).Although the malignant degree of LGG is low and the survival prognosis of patients after surgical treatment is good,LGG is still invasive and can progress to glioblastoma.At present,there is no consensus on the treatment strategy of LGG.Some studies believe that factors such as the scope of surgical resection and the age of patients are risk factors for patients with LGG,but neoadjuvant therapy such as radiotherapy and chemotherapy is still controversial.Therefore,a large sample of clinical data is needed to analyze the independent prognostic risk factors of patients with LGG.Methods:The clinical data of 3638 LGG patients from the SEER(The Surveillance,Epidemiology,and End Results)database of the National Cancer Institute of the United States were collected,including sex,age,race,marital status,pathological diagnosis type,primary location of tumor,laterality,number of tumors,tumor size,mode of operation,radiotherapy and chemotherapy.First of all,using the method of random grouping,the cases included in the study(n=3638)were divided into training set(n=2548)and verification set(n=1090).Univariate and multivariate Cox regression and Lasso regression were used to analyze the independent prognostic risk factors of LGG patients,and Kaplan-Meier survival curve was used to observe the survival rate of LGG patients in different groups.Cox regression prognostic risk model was constructed.Prognostic factors were divided into high-risk group and low-risk group as variables.ROC curve and Kaplan-Meier curve were used to evaluate the clinical predictive value of risk score.Finally,the independent prognostic risk factors of LGG patients were visually analyzed to draw Nomogram diagrams to further predict the effects of prognostic risk factors on the 3-,5-and 10-year survival rates of LGG patients.The reliability and accuracy of the predictive model were evaluated by consistency index and Calibration curves.Results:1.Univariate Cox analysis showed that age,marital status,pathological type,primary location of tumor,laterality,number of tumors,tumor size,mode of operation,radiotherapy and chemotherapy were significantly correlated with the prognosis of LGG patients in training set and verification set(p < 0.05).The prognosis of LGG patients was significantly correlated with age,marital status,pathological type,primary location,laterality,number of tumors,tumor size,mode of operation,radiotherapy and chemotherapy.2.KM survival curve analysis showed that there were significant differences in age,marital status,pathological type,primary location of tumor,laterality,number of tumors,tumor size,mode of operation,radiotherapy and chemotherapy in the training group(p < 0.001),and verified in the verification set(p < 0.001).3.Lasso-cox regression analysis showed that patients’ age,marital status,pathological type,primary location of tumor,number of tumors,tumor size,mode of operation,radiotherapy and chemotherapy could be independent prognostic risk factors of low-grade glioma.However,chemotherapy as an independent risk factor in the training set was statistically significant(p < 0.001),but in the verification set(p >0.05).4.The prognostic risk model was established by multivariate Cox regression analysis,and the ROC curve analysis risk score was used to predict the prognosis of LGG patients.The results showed that the AUC values of 5 and 10 years in the training set were 0.841,0.809 and 0.789,respectively,and the verification results in the verification set were 0.845,0.818 and 0.789,respectively.The results of survival curve analysis between high-risk and low-risk groups of LGG patients in training set and verification set were all p < 0.001.5.Nine independent prognostic risk factors and risk scores of LGG patients were included to construct the Nomogram model.The accuracy of the model in predicting the overall survival time of patients was high,and the C-index in the training set and verification set was 0.776 and 0.785 respectively.Calibration curve analysis showed that there was a high degree of fit between the predicted survival rate of Nomogram and the actual survival rate of patients with LGG.Conclusion:1.This study shows that age,marital status,pathological type,primary location of tumor,number of tumors,tumor size,mode of operation,radiotherapy and chemotherapy are independent prognostic risk factors for patients with LGG.2.In the Lasso-cox prognostic classification model,LGG patients in the lowrisk group had survival benefits,and the survival rate was higher than that in the high-risk group.3.The Nomogram prognostic model constructed in this study to predict the overall survival time of LGG patients has strong predictive ability and good clinical application value,and can be used as an important reference for individualized treatment of LGG patients.