Study On Survival Prediction Model Based On Survival Analysis Of Tongue Cancer In Hunan

Objective:(1)To estimate the median survival time and t-year(t=1,3,5)survival rate of patients with tongue cancer and to screen out the protective and risk factors affecting the prognosis of patients with tongue cancer.To construct the individual prognostic index equation,Bayes discriminant model and line diagram model of patients with tongue cancer,(2)to provide reference for prognosis prediction and treatment choice of patients with tongue cancer.Methods:(1)Based on literature and consulting experts,a collection table of general characteristics,clinical data and follow-up data of tongue cancer patients was designed.(2)According to the data collection table,the general situation and Clinical characteristics of tongue cancer patients who were diagnosed by pathology in the medical record management system of one Cancer Hospital and were first hospitalized in one Cancer Hospital from January 1,2010 to December 31,2016 were collected.The survival rate of patients with different characteristics of tongue cancer was estimated and compared by Kaplan-Meier method and Log-rank method.On the basis of the results of univariate analysis,the multivariate Cox regression method was used to determine the factors affecting the survival of patients with tongue cancer,without special explanation,all the hypothesis test level was 0.05,bilateral;(3)According to the results ofmulti-factor analysis,the individual prognostic index equation,Bayes discriminant model,and the predictive ability of the two models were evaluated synthetically,and the application value was judged by the probability of misjudgment.The above contents are completed in the statistical software SPSS20.0.The software Medcalc was used to compare the prediction capabilities of the two models.(4)the nomogram is constructed with the rms package in R software 3.5.3 to predict the effect used consistency index for the evaluation.Results:(1)A total of 818 patients were followed up in this study.The number of people who died of tongue cancer was 274,the number of survivors was 474,the number of missing visitors was 70,and the rate of missing visits was 8.1%.Patients with tongue cancer had a median survival time of 5.50 years(95%CI:5.359,5.475),a one-year survival rate of 99.9%(95%CI:99.7,99.9),a 3-year survival rate of 99.0%(95%CI:98.7,99.3)and a 5-year survival rate of 84.7%(95%CI: 84.3,85.1).(2)Single factor analysis showed: tongue cancer location,transfer,neck lymph node enlargement,the range of lesions,treatments,skin flap repair,neck lymph node cleaning,anesthesia method,level of incision healing,chemotherapy,case classification,hospitalization days,histological grade,TNM stage,when the discharge status may be the influence factors of patients with tongue cancer survival(all p<0.05)(3)Multivariate Cox regression analysis results showed that the cervical lymph nodeenlargement,lymph node metastasis,the range of lesions,surgical anesthesia,histological grading,discharge condition when the six variables are the factors influencing patients with tongue cancer survival time(p < 0.05),cervical lymph node metastases,neck lymph nodes and swelling,lesion range greater than 6 cm,intraoperative anesthesia without anesthesia,low degree of pathological differentiation and low discharge state is a risk factor for patients with tongue cancer survival;(4)The evaluation results of individual prognostic index equation prognosis prediction capabilities show that the area under the ROC curve is 0.615,the sensitivity of the prognosis index equation is 59.6%,the specificity is 60.9%,and the probability of misjudgment is 0.39;the evaluation result of the prognosis prediction ability of the Bayes discriminant model shows: Bayes discrimination The sensitivity of the model is 68.3%,the specificity is 56.7%,the probability of misjudgment is 0.34,the area under the ROC curve is 0.615,p=0.002.There is a difference in the area under the ROC curve between the individual prognostic index equation model and the Bayes discriminant model.(5)nomogram prediction ability evaluation,the results show that the consistency index is 0.61,in the range of 0.5-0.7,the model has some credibility to evaluate the prognosis of patients.ConClusions:(1)Whether or not the cervical lymph nodes are enlarged,accompanied by lymph node metastasis,the range of lesions,surgical anesthesia,the degree of pathological differentiation,and the condition at the time of discharge are the influencing factors for the survival of patients with tongue cancer.The patients with tongue cancer have longer survival time when the lymph nodes are not enlarged,no lymph node metastasis occurs,the range of lesions is small,the anesthesia mode is general anesthesia at the time of operation,the higher degree of pathological differentiation and the condition is good at the time of discharge.(2)The individual prognostic index equation,Bayes discriminant model and nomogram model of tongue cancer patients constructed by this research have reference value for predicting the prognosis of tongue cancer patients,and can provide a reference basis for clinical prognosis assessment of tongue cancer patients.The model can be more simple and intuitive to predict the prognosis of tongue cancer patients.(3)Compared with the individual prognostic index equation,the prediction effect of Bayes discriminant model has higher accuracy.