Risk factors for Drug‐resistant Epilepsy (DRE) and a nomogram model to predict DRE development in post‐traumatic epilepsy patients

Abstract Objectives To identify factors affecting the development of drug‐resistant epilepsy (DRE), and establish a reliable nomogram to predict DRE development in post‐traumatic epilepsy (PTE) patients. Methods This study conducted a retrospective clinical analysis in patients with PTE who visited the Epilepsy Center, Beijing Tiantan Hospital from January 2013 to December 2018. All participants were followed up for at least 3 years, and the development of DRE was assessed. Data from January 2013 to December 2017 were used as development dataset for model building. Those independent predictors of DRE were included in the final multivariable logistic regression, and a derived nomogram was built. Data from January 2018 to December 2018 were used as validation dataset for internal validation. Results Complete clinical information was available for 2830 PTE patients (development dataset: 2023; validation dataset: 807), of which 21.06% (n = 596) developed DRE. Among all parameters of interest including gender, age at PTE, family history, severity of traumatic brain injury (TBI), single or multiple injuries, lesion location, post‐TBI treatments, acute seizures, PTE latency, seizure type, status epilepticus (SE), and electroencephalogram (EEG) findings, four predictors showed independent effect on DRE, they were age at PTE, seizure type, SE, and EEG findings. A model incorporating these four variables was created, and a nomogram to calculate the probability of DRE using the coefficients of the model was developed. The C‐index of the predictive model and the validation was 0.662 and 0.690, respectively. The goodness‐of‐fit test indicated good calibration for model development and validation (p = 0.272, 0.572). Conclusions The proposed nomogram achieved significant potential for clinical utility in the prediction of DRE among PTE patients. The risk of DRE for individual PTE patients can be estimated by using this nomogram, and identified high‐risk patients might benefit from non‐pharmacological therapies at an early stage.


| INTRODUC TI ON
Traumatic brain injury (TBI) is a global public health concern, with high morbidity, disability and fatality rate. 1 Post-traumatic seizure (PTS) and post-traumatic epilepsy (PTE, defined as recurrent unprovoked PTS occurs more than 7 days after TBI) are common complications and sequelae after TBI. 2 Early PTS also known as acute seizure, refers to seizures that occur within 7 days after TBI. Early intervention with antiseizure medications (ASMs) can effectively prevent acute seizure after TBI; however, it cannot prevent the development of PTE. 3,4 Most of the patients with PTE have a good prognosis, but there are still some patients who have poor responses to standard ASMs therapy and develop drugresistant epilepsy (DRE) with frequent seizures. 2,5 According to the numbers of published researches, the severity of TBI is a wellestablished risk factor for PTE. [6][7][8] Demographic factors (such as male gender, age ≥65 years), 9,10 specific TBI type and TBI characteristics (such as penetrating injury, skull fracture, hemorrhagic lesion, temporal lobe trauma, prolonged loss of consciousness, and post-traumatic amnesia) 2, 6,[9][10][11] are also associated with the development of PTE after TBI. It is reported that PTE accounts for more than 5% of all patients with DRE transferred to epilepsy centers for surgical evaluation, 12 and it takes a tremendous toll on patients both physically, mentally, and brings significant economic and mental burden to the family and society. Early prophylactic ASM therapy after TBI showed no protective effect on reducing the incidence of PTE. [13][14][15] What is more, with the increase of ASM treatment options, the response of epilepsy patients to drugs gradually decreased, 16 17 It is imperative to identify risk factors for DRE and construct an accurate prognostic model to predict DRE development by longitudinally assessing a well-powered sample of PTE patients.
A nomogram is a graphic score providing individualized predictions for clinical outcomes, such as the emergence of a disease or death. 18 Nomograms are widely used for cancer prognosis, 18,19 and in recent years some researchers have also used them in clinical studies of neurological diseases. [20][21][22]

| Data collection
Data were collected through "the PTE patient information registration form" mentioned in our previously published study. 24 In addition to demographic information, family history, personal medical records, TBI details, the clinical condition of PTE (including the presence of acute seizure, latency of PTE, the type and frequency of seizure), and the electroencephalogram (EEG), all patients were followed for at least 3 years. Close attention was paid to the usage of ASMs and the drug response of individual PTE patients. The response of ASMs was evaluated according to the seizure type and frequency before and after using appropriately chosen ASMs by two experienced neurologists (TTY and QW).
Once the patient met the criteria of 2009 definition of DRE by ILAE: "failure of achieving a seizure-free duration of 3 times the interseizure interval or 1 year (depending on which is longer) of two tolerated, appropriately chosen and used antiepileptic drugs (whether as monotherapies or in combination)", 25 he/she was assessed as developing DRE.

K E Y W O R D S
drug-resistant epilepsy, nomogram, post-traumatic epilepsy, prognosis, risk factor Concerning TBI, the severity was evaluated based on neurological and imaging evaluations. 6 "Severe TBI" is characterized by one or more of the following features: brain contusion, intracranial hematoma, or loss of consciousness or post-traumatic amnesia lasting ≥24 h. Otherwise, TBI was evaluated as "mild-to-moderate TBI". This study also recorded the lesion location, and accessed the craniocerebral injury as a single injury or multiple injuries according to lesion caused by the TBI (single injury: a single or continuous lesion, for example, unilateral frontotemporal; multiple injuries: lesions of bilateral involvement or topographically separate locations, for example, bilateral frontal). 24 Moreover, post-TBI treatments were recorded, According to the description of the patient or family members and the medical record, we evaluated whether the patient had status epilepticus (SE). The reference standard used was the 2015 classification of SE by ILAE. 27 The type of seizure and the presence of SE was assessed solely based on the condition of the first 2 years of PTE course, the condition of 2 years later was not considered.
By reviewing the original EEG data or the report of EEG, this study recorded the EEG as "normal EEG", "abnormal background without epileptiform discharges", or "epileptiform discharges". The EEG of all patients was any one of the interictal EEG during outpatient visit, and the duration of monitoring was 20-40 min.
Between September 2019 and August 2021, all patients were monitored in clinic or by telephone. All patients were followed continuously for at least 3 years before the last follow-up. Modified Rankin Scale (mRS) score and the development of DRE were assessed at that time. Two neurologists (TTY and QW) participated in the assessment of the drug resistance status of the individual PTE patient. SPSS 23.0 software (IBM Crop.) and R version 4.1.1 software were used for data analysis. Numerical data were represented by percentages, and continuous data were represented by mean ± standard deviation (SD) or median and interquartile range (IQR). The χ 2 or Fisher exact test was used to compare numerical data as appropriate, and Mann-Whitney U test was used to compare continuous data. A two-sided p < 0.05 was deemed significant. Univariate and multivariate logistic regression were used to identify factors have an impact on DRE.

| Patient characteristics
The clinical data of all patients diagnosed with "PTE" at our epilepsy center between January 2013 and December 2018 were reviewed, and 3042 patients who met the inclusion criteria mentioned above were screened. 212 patients were excluded because of the exclusion criteria mentioned above, data from 2830 patients with PTE were finally included in this study for analysis ( Figure 1).

| Nomogram model development and validation
A model incorporating these four characteristics was created according to the multiple logistic regression results, and a nomogram to calculate the probability of DRE using the coefficients of the model was developed (Figure 2). In this nomogram, the predictor of age at PTE was assigned of 100 points, followed by the presence of SE, and the EEG findings had the least effect on DRE development.
This nomogram provides convenience when predicting the probabil-   potential resection operations, and laser-induced thermal therapy. 35 Numerous studies have reported that VNS effectively reduce seizure frequency in DRE patients as an adjunctive therapy, 36 it was also reported that VNS can cause spontaneous neural activity changes in an ongoing process. 37 DBS in the appropriate stimulation targets (such as medial septum for temporal lobe epilepsy) was also reported not only reduced spontaneous seizures, but also improved behavioral performance. 38 These results suggest that if PTE patients at high risk of DRE could be identified and treated with non-pharmacological therapies in the early stage, they might benefit early, and non-pharmacological therapies in the early stage might help reduce the physical and mental damage of repeated seizures, improve their quality of life. 35 This also underscores the importance of identifying patients at high risk for DRE.

| DISCUSS ION
The results of this study indicated that demographic characteristics played roles in predicting the progression of PTE patients to DRE. We found that female patients tended to be more likely to develop DRE than male patients (though the difference was not statistically significant). This was consistent with previous findings that females were at a higher risk for PTE after TBI. 7  Using a predictive tool helps identify PTE patients at a high risk of developing DRE, who might need closer and more regular monitoring. We physicians might make a more individualized, aggressive plan (for example, VNS, DBS, or resection operations) for those patients at the early stage of PTE.
As a retrospective study, we also realized that this study has several limitations. First, there might be some unobserved and/ or uncontrolled confounding factors and we might miss a few factors affecting the development of DRE. Second, though we tried to make our assessment of the drug resistance status reliable, it is undeniable that some "DRE patients" may have "pseudo-resistance". 49 Third, the data of several variables were based on physician reports or descriptions of patients or family members; therefore, may lead to possible information bias. Further population-based prospective studies are needed to clarify the risk factors for DRE among PTE patients fully.  ditor.com) for the English language editing and review services.

CO N FLI C T O F I NTE R E S T
None of the authors has any conflict of interest to disclose.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the funding of this study are available from the corresponding author upon reasonable request.

I N FO R M ED CO N S ENT
The study was conducted in accordance with the Declaration of Helsinki, and all participants provided informed consent for the use of their medical records.