Development And Validation Of Prediction Model For Seizure Freedom With AED Treatment In Newly Diagnosed Patients With MRI-negative Epilepsy

Objective:Anti-epileptic drugs(AEDs)are the basic choice in newly diagnosed patients with MRI-negative epilepsy;however,the response is huge different between individuals,even in those with the same type of epilepsy or seizures,and 30-40%of patients will develop drug-resistant epilepsy,and selection of personalized AED for a newly diagnosed patients with epilepsy has become a hot-spot.Magnetic resonance imaging(MRI)-negative epilepsy is a huge challenge for surgical treatment because there is no clear anatomic localization of epileptogenic focus and AEDs are the primary treatment.However,surgery may be more suitable for some patients with MRI-positive epilepsy.Although several factors assessing the efficacy to AEDs have been identified,there is lack of a comprehensive prediction model for AED treatment and to make clinical decision in patients with MRI-negative epilepsy.Nomograms are commonly used to predict prognosis by integrating multiple variables into a graphical tool.Several nomograms are currently employed in epilepsy,including the prediction of seizure outcomes after epilepsy surgery,prediction of seizure recurrence after withdrawal of AEDs and prediction of seizures in pregnant women with epilepsy.Thus,the aim of our study was to identify the predictive factors that affect the clinical outcome on AEDs and to develop and validate a personalized model based on nomogram to predict the probability of 6-month and 12-month seizure freedom with AEDs in newly diagnosed patients with MRI-negative epilepsy,which may be designed to guide the decision-making.Methods:The development cohort of newly diagnosed patients with MRI-negative epilepsy at the Epilepsy Center of Henan Provincial People's Hospital from 2015 to 2018 were retrospective enrolled.Clinical and electroencephalogram(EEG)features were extracted.Univariate analysis and a forward and backward elimination of multivariate Cox regression analysis were used to select predictive factors.A prediction model was developed based on nomogram.The model was externally validated at two independent epilepsy centers in the Second Affiliated Hospital of Xinxiang Medical University and the First Affiliated Hospital of Henan University.Predictive accuracy was assessed using discrimination and calibration,and the clinical values were measured by decision curve analysis.The total scores were categorized into three levels using a classification and regression tree model.A web-based calculator was also established for convenient application.Results:The development cohort included 543 consecutive newly diagnosed patients with MRI-negative epilepsy patients,while the validation cohorts included 493 in two independent cohorts.Nineteen factors were extracted and the factors of prediction of poor possibility to AEDs included circadian rhythms,EEG before AEDs,neuropsychiatric disorders,perinatal brain injury and history of central nervous system infection in the final model.The model showed good discrimination with C-index(development group:0.83;validation group 0.78).Calibration plots indicated high agreement of predicted and observed outcomes.The decision curve analysis showed good-predictive accuracy and clinical values.The points of the score can be categorized to three levels:high probability(0-83.11 points),medium probability(83.11-122.71)points and low probability(>122.71 points).A web-based calculator was established to make this model widely available and easily applicable in clinical practice.Conclusions:We identified poor predictors to AEDs and established and validated a simple,practical,evidence-based prediction tool for the possibility of seizure freedom to AED treatment to aid in the treatment decision for newly diagnosed patients with MRI-negative epilepsy,which can also be for clinical consultation and study design.