Nomogram For Predicting Early Postoperative Seizures In Patients After Craniocerebral Tumor Surgery

Objective Early Postoperative epilepsy(EPS)is a common complication after craniocerebral tumors.The occurrence of EPS can aggravate the injury of brain tissue,affect the prognosis of patients,and aggravate the postoperative condition of patients.The purpose of this study was to construct a EPS risk prediction model for patients with craniocerebral tumor and evaluate the prediction efficiency,differentiation,calibration and clinical applicability of the model.Methods Patients who underwent craniotomy for tumor resection in the neurosurgery department of a grade III class A hospital in Anhui Province from January 2019 to April 2021 were selected as the study subjects,and 402 patients were included in the study,who were divided into sezures group 42 cases and non sezures group 360 cases on the basis of whether the patients had sezures within one week after surgery.T-test and chi-square test was used to analyze the influence of each variable on the occurrence of EPS.The differences were statistically significant(P<0.05)were incorporated into binary Logistic regression to screen out the independent influencing factors of EPS occurrence(P<0.05).Then the varivables were introduced into R software(R 3.3.2),and the RMS program package was used to establish the prediction model of nomogram.Further,non-parametric Bootstrap resampling times=1000 times was applied to verify and evaluate the model internally.The differentiation degree of the line graph model was evaluated by the area under ROC curve(AUC ROC)of the line graph model before and after internal verification.The calibration degree of the line graph model was evaluated by using hosmer-Lemeshow goodness of fit test and drawing calibration curves.The clinical applicability of the rosette model was evaluated by clinical decision curve(DCA).Results A total of 402 patients undergoing intracranial tumor resection were investigated in this study,among which 42 patients developed in EPS(10.4%).Univariate analysis showed that the two groups had preoperative history of epilepsy,tumor location,surgical approach,tumor diameter,tumor resection degree,postoperative edema and postoperative bleeding.There were statistically significant differences in terms of item factors(P<0.05).These 7 factors were included in binary Logistic regression analysis,and the results showed that preoperative history of epilepsy(OR= 2.870,95%CI:1.131-7.283),tumor resection degree(OR=0.337,95%CI: 0.127-0.897),postoperative edema in the operative area(OR=2.405,95%CI:1.102-5.250),postoperative bleeding(OR=3.143,95%CI:1.126-8.125)and the tumor location(OR=0.028,95%CI:0.001-0.524)were independent influencing factors of early epilepsy after craniocerebral tumor surgery,and the difference was statistically significant(P<0.05).Hence the above 5 factors were used to construct the nomogram.(1)Harrell’s C-index of the nomogram was 0.821(95%CI: 0.754-0.888),the sensitivity was 0.857,and the specificity was 0.625,suggesting that the nomogram has high prediction efficiency and differentiation.(2)And internal verification of the nomogram was carried out by Bootstrap self-sampling 1000 times,showing that the area under ROC curve of the nomogram after internal verification was 0.831(95%CI: 0.765-0.886),which suggests it has high prediction efficiency and stability.(3)Hosmer-leweshow deviation test showed that there was no statistical significance in the deviation between the predicted probability and the actual frequency of occurrence of the line graph(χ2=4.253,P=0.643).Internal verification of the rosette model was carried out by computer simulation of self-sampling(Bootstap,1000 times of sampling).According to the calibration curve of the rosette model before and after internal verification,the Mean absolute error between the predicted probability and the actual occurrence frequency of the rosette model before and after internal verification(Mean Absolute error)Error and MAE are 0.023 and 0.019 respectively,indicating that the lipopograph has a good calibration degree.(4)Using the “RMDA” program package to draw the DCA of the nomogram,it was found that when the risk threshold of early postoperative epilepsy was 0.028-0.960,the clinical net return of using the rosette model was significantly higher than that of “full intervention”and “no intervention”,suggesting that the rosette model has better clinical applicability.Conclusion The five factors,preoperative history of epilepsy,degree of tumor resection,postoperative edema,postoperative bleeding and tumor location,are independent influencing factors for the occurrence of early postoperative epilepsy in patients with craniocerebral tumor.And the prediction model of nomogram constructed based on the above factors has a high degree of calibration and consistency.