Identification of patients who will not achieve seizure remission within 5 years on AEDs

Predicting efficacy of antiseizure medication treatment with machine learning algorithms in North Indian population

PURPOSE

This study aimed to develop a classifier using supervised machine learning to effectively assess the impact of clinical, demographical, and biochemical factors in accurately predicting the antiseizure medications (ASMs) treatment response in people with epilepsy (PWE).

METHODS

Data was collected from 786 PWE at the Outpatient Department of Neurology, Institute of Human Behavior and Allied Sciences (IHBAS), New Delhi, India from 2005 to 2015. Patients were followed up at the 2nd, 4th, 8th, and 12th month over the span of 1 year for the drugs being administered and their dosage, the serum drug levels, the frequency of seizure control, drug efficacy, the adverse drug reactions (ADRs), and their compliance to ASMs. Several features, including demographic details, medical history, and auxiliary examinations electroencephalogram (EEG) or Computed Tomography (CT) were chosen to discern between patients with distinct remission outcomes. Remission outcomes were categorized into 'good responder (GR)' and 'poor responder (PR)' based on the number of seizures experienced by the patients over the study duration. Our dataset was utilized to train seven classical machine learning algorithms i.e Extreme Gradient Boost (XGB), K-Nearest Neighbor (KNN), Support Vector Classifier (SVC), Decision Tree (DT), Random Forest (RF), Naïve Bayes (NB) and Logistic Regression (LR) to construct classification models.

RESULTS

Our research findings indicate that 1) among the seven algorithms examined, XGB and SVC demonstrated superior predictive performances of ASM treatment outcomes with an accuracy of 0.66 each and ROC-AUC scores of 0.67 (XGB) and 0.66 (SVC) in distinguishing between PR and GR patients. 2) The most influential factor in discerning PR to GR patients is a family history of seizures (no), education (literate) and multitherapy with Chi-square (χ2) values of 12.1539, 8.7232 and 13.620 respectively and odds ratio (OR) of 2.2671, 0.4467, and 1.9453 each. 3). Furthermore, our surrogate analysis revealed that the null hypothesis for both XGB and SVC was rejected at a 100 % confidence level, underscoring the significance of their predictive performance. These findings underscore the robustness and reliability of XGB and SVC in our predictive modelling framework.

SIGNIFICANCE

Utilizing XG Boost and SVC-based machine learning classifier, we successfully forecasted the likelihood of a patient's response to ASM treatment, categorizing them as either PR or GR, post-completion of standard epilepsy examinations. The classifier's predictions were found to be statistically significant, suggesting their potential utility in improving treatment strategies, particularly in the personalized selection of ASM regimens for individual epilepsy patients.

The pharmacological treatment of epilepsy in adults

The pharmacological treatment of epilepsy entails several critical decisions that need to be based on an individual careful risk-benefit analysis. These include when to initiate treatment and with which antiseizure medication (ASM). With more than 25 ASMs on the market, physicians have opportunities to tailor the treatment to individual patients´ needs. ASM selection is primarily based on the patient's type of epilepsy and spectrum of ASM efficacy, but several other factors must be considered. These include age, sex, comorbidities, and concomitant medications to mention the most important. Individual susceptibility to adverse drug effects, ease of use, costs, and personal preferences should also be taken into account. Once an ASM has been selected, the next step is to decide on an individual target maintenance dose and a titration scheme to reach this dose. When the clinical circumstances permit, a slow titration is generally preferred since it is associated with improved tolerability. The maintenance dose is adjusted based on the clinical response aiming at the lowest effective dose. Therapeutic drug monitoring can be of value in efforts to establish the optimal dose. If the first monotherapy fails to control seizures without significant adverse effects, the next step will be to gradually switch to an alternative monotherapy, or sometimes to add another ASM. If an add-on is considered, combining ASMs with different modes of action is usually recommended. Misdiagnosis of epilepsy, non-adherence and suboptimal dosing are frequent causes of treatment failure and should be excluded before a patient is regarded as drug-resistant. Other treatment modalities, including epilepsy surgery, neuromodulation, and dietary therapies, should be considered for truly drug-resistant patients. After some years of seizure freedom, the question of ASM withdrawal often arises. Although successful in many, withdrawal is also associated with risks and the decision needs to be based on careful risk-benefit analysis.

Association Between Angiotensin Receptor Blocker Therapy and Incidence of Epilepsy in Patients With Hypertension

Importance

Arterial hypertension is associated with an increased incidence of epilepsy. Results from animal studies suggest that angiotensin receptor blocker (ARB) therapy could inhibit epileptic seizures. However, there is a lack of clinical data to support the use of ARB therapy in humans.

Objective

To assess whether ARB therapy is associated with a decreased incidence of epilepsy in patients with hypertension.

Design, Setting, and Participants

This cohort study obtained data from the Disease Analyzer database (IQVIA) on patients aged 18 years or older who had hypertension and at least 1 antihypertensive drug prescription. Patients were treated at 1274 general practices between January 2010 and December 2020 in Germany. Data were available for 1 553 875 patients who had been prescribed at least 1 antihypertensive drug. Patients diagnosed with epilepsy before or up to 3 months after the index date were excluded. A total of 168 612 patients were included in propensity score matching. Patients treated with 1 of 4 antihypertensive drug classes (β-blockers, ARBs, angiotensin-converting enzyme inhibitors, and calcium channel blockers [CCBs]) were matched to each other using propensity scores.

Main Outcomes and Measures

The main outcome of the study was the incidence of epilepsy associated with ARB therapy compared with other antihypertensive drug classes. Cox regression models were used to study the association between the incidence of epilepsy and ARBs compared with all other antihypertensive drug classes as a group.

Results

The study included a total of 168 612 patients, with 42 153 in each antihypertensive drug class. The mean [SD] age of patients was 62.3 [13.5] years, and 21 667 (51.4%) were women. The incidence of epilepsy within 5 years was lowest among patients treated with ARBs (0.27% at 1 year, 0.63% at 3 years, 0.99% at 5 years) and highest among patients receiving β-blockers and CCBs (0.38% for both β-blockers and CCBs at 1 year; 0.91% for β-blockers and 0.93% for CCBs at 3 years; β-blockers, 1.47%; and CCBs, 1.48% at 5 years). Angiotensin receptor blocker therapy was associated with a significantly decreased incidence of epilepsy (hazard ratio, 0.77; 95% CI, 0.65-0.90) compared with the other drug classes as a group.

Conclusions and Relevance

In this cohort study of patients with hypertension, ARB therapy was associated with a significantly decreased incidence of epilepsy. The findings suggest antihypertensive drugs could be used as a novel approach for preventing epilepsy in patients with arterial hypertension.

Machine learning models for decision support in epilepsy management: A critical review

PURPOSE

There remain major challenges for the clinician in managing patients with epilepsy effectively. Choosing anti-seizure medications (ASMs) is subject to trial and error. About one-third of patients have drug-resistant epilepsy (DRE). Surgery may be considered for selected patients, but time from diagnosis to surgery averages 20 years. We reviewed the potential use of machine learning (ML) predictive models as clinical decision support tools to help address some of these issues.

METHODS

We conducted a comprehensive search of Medline and Embase of studies that investigated the application of ML in epilepsy management in terms of predicting ASM responsiveness, predicting DRE, identifying surgical candidates, and predicting epilepsy surgery outcomes. Original articles addressing these 4 areas published in English between 2000 and 2020 were included.

RESULTS

We identified 24 relevant articles: 6 on ASM responsiveness, 3 on DRE prediction, 2 on identifying surgical candidates, and 13 on predicting surgical outcomes. A variety of potential predictors were used including clinical, neuropsychological, imaging, electroencephalography, and health system claims data. A number of different ML algorithms and approaches were used for prediction, but only one study utilized deep learning methods. Some models show promising performance with areas under the curve above 0.9. However, most were single setting studies (18 of 24) with small sample sizes (median number of patients 55), with the exception of 3 studies that utilized large databases and 3 studies that performed external validation. There was a lack of standardization in reporting model performance. None of the models reviewed have been prospectively evaluated for their clinical benefits.

CONCLUSION

The utility of ML models for clinical decision support in epilepsy management remains to be determined. Future research should be directed toward conducting larger studies with external validation, standardization of reporting, and prospective evaluation of the ML model on patient outcomes.

Altered static and dynamic functional network connectivity in temporal lobe epilepsy with different disease duration and their relationships with attention

The brain network alterations associated with temporal lobe epilepsy (TLE) progression are still unclear. The purpose of this study was to investigate altered patterns of static and dynamic functional network connectivity (sFNC and dFNC) in TLE with different durations of disease. In this study, 19 TLE patients with a disease duration of ≤5 years (TLE-SD), 24 TLE patients with a disease duration of >5 years (TLE-LD), and 21 healthy controls (HCs) underwent resting-state functional magnetic resonance imaging and attention network test. We used group independent component analysis to determine the target resting-state networks. Sliding window correlation and k-means clustering analysis methods were used to obtain different dFNC states, temporal properties, and temporal variability. We then compared sFNC and dFNC between groups and found that compared with HCs, TLE-SD patients had increased sFNC between the dorsal attention network and sensorimotor network/visual network (VN), but decreased sFNC between the inferior-posterior default mode network and VN. In the strongly connected dFNC state, TLE-SD patients spent more time, had greater mean dwell time, and showed greater inconsistent abnormal network connectivity. There was a significant negative correlation between the temporal variability of auditory network- left fronto-parietal network connectivity and orienting effect. No significant differences in sFNC and dFNC were detected between TLE-LD and HC groups. These findings suggest that the damage and functional brain network abnormalities gradually occur in TLE patients after the onset of epilepsy, which might lead to functional network reorganization and compensatory remodeling as the disease progresses.

A population-based study to develop juvenile arthritis case definitions for administrative health data using model-based dynamic classification

BACKGROUND

Previous research has shown that chronic disease case definitions constructed using population-based administrative health data may have low accuracy for ascertaining cases of episodic diseases such as rheumatoid arthritis, which are characterized by periods of good health followed by periods of illness. No studies have considered a dynamic approach that uses statistical (i.e., probability) models for repeated measures data to classify individuals into disease, non-disease, and indeterminate categories as an alternative to deterministic (i.e., non-probability) methods that use summary data for case ascertainment. The research objectives were to validate a model-based dynamic classification approach for ascertaining cases of juvenile arthritis (JA) from administrative data, and compare its performance with a deterministic approach for case ascertainment.

METHODS

The study cohort was comprised of JA cases and non-JA controls 16 years or younger identified from a pediatric clinical registry in the Canadian province of Manitoba and born between 1980 and 2002. Registry data were linked to hospital records and physician billing claims up to 2018. Longitudinal discriminant analysis (LoDA) models and dynamic classification were applied to annual healthcare utilization measures. The deterministic case definition was based on JA diagnoses in healthcare use data anytime between birth and age 16 years; it required one hospitalization ever or two physician visits. Case definitions based on model-based dynamic classification and deterministic approaches were assessed on sensitivity, specificity, and positive and negative predictive values (PPV, NPV). Mean time to classification was also measured for the former.

RESULTS

The cohort included 797 individuals; 386 (48.4 %) were JA cases. A model-based dynamic classification approach using an annual measure of any JA-related healthcare contact had sensitivity = 0.70 and PPV = 0.82. Mean classification time was 9.21 years. The deterministic case definition had sensitivity = 0.91 and PPV = 0.92.

CONCLUSIONS

A model-based dynamic classification approach had lower accuracy for ascertaining JA cases than a deterministic approach. However, the dynamic approach required a shorter duration of time to produce a case definition with acceptable PPV. The choice of methods to construct case definitions and their performance may depend on the characteristics of the chronic disease under investigation.

Genetic factors and the risk of drug-resistant epilepsy in young children with epilepsy and neurodevelopment disability: A prospective study and updated meta-analysis

Drug-resistant epilepsy (DRE) affects 7% to 20% of children with epilepsy. Although some risk factors for DRE have been identified, the results have not been consistent. Moreover, data regarding the risk factors for epilepsy and its seizure outcome in the first 2 years of life are limited.We analyzed data for children aged 0 to 2 years with epilepsy and neurodevelopmental disability from January, 2013, through December, 2017. These patients were followed up to compare the risk of DRE in patients with genetic defect (genetic group) with that without genetic defect (nongenetic group). Additionally, we conducted a meta-analysis to identify the pooled prevalence of genetic factors in children with DRE.A total of 96 patients were enrolled. A total of 68 patients were enrolled in the nongenetic group, whereas 28 patients were enrolled in the genetic group. The overall DRE risk in the genetic group was 6.5 times (95% confidence interval [CI], 2.15-19.6; p = 0.03) higher than that in the nongenetic group. Separately, a total of 1308 DRE patients were participated in the meta-analysis. The pooled prevalence of these patients with genetic factors was 22.8% (95% CI 17.4-29.3).The genetic defect plays a crucial role in the development of DRE in younger children with epilepsy and neurodevelopmental disability. The results can serve as a reference for further studies of epilepsy panel design and may also assist in the development of improved treatments and prevention strategies for DRE.

Brain angiotensin system: a new promise in the management of epilepsy?

Epilepsy is a highly prevalent neurological disease and anti-epileptic drugs (AED) are almost the unique clinical treatment option. A disbalanced brain renin-angiotensin system (RAS) has been proposed in epilepsy and several reports have shown that angiotensin II (Ang II) receptor-1 (ATR1) activation is pro-inflammatory and pro-epileptogenic. In agreement, ATR1 blockage with the repurposed drug losartan has shown benefits in animal models of epilepsy. Processing of Ang II by ACE2 enzyme renders Ang-(1-7), a metabolite that activates the mitochondrial assembly (Mas) receptor (MasR) pathway. MasR activation presents beneficial effects, facilitating vasodilatation, increasing anti-inflammatory and antioxidative responses. In a recent paper published in Clinical Science, Gomes and colleagues (Clin. Sci. (Lond.) (2020) 134, 2263-2277) performed intracerebroventricular (icv) infusion of Ang-(1-7) in animals subjected to the pilocarpine model of epilepsy, starting after the first spontaneous motor seizure (SMS). They showed that this approach reduced the frequency of SMS, restored animal anxiety, increased exploration, and augmented the hippocampal expression of protective catalase enzyme and antiapoptotic protein B-cell lymphoma 2 (Bcl-2). Interestingly, but surprisingly, Gomes and colleagues showed that MasR expression and mTor activity were reduced in the hippocampus of the epileptic Ang-(1-7) treated animals. These results show that Ang-(1-7) administration could represent a new avenue for developing strategies for the management of epilepsy in clinical settings. However, future work is necessary to evaluate the levels of RAS metabolites and the activity of key enzymes in these experimental interventions to completely understand the therapeutic potential of the brain RAS manipulation in epilepsy.

Efficacy of adjunctive cenobamate based on number of concomitant antiseizure medications, seizure frequency, and epilepsy duration at baseline: A post-hoc analysis of a randomized clinical study

BACKGROUND

In an 18-week, double-blind, placebo-controlled study (YKP3089C017; NCT01866111), cenobamate was effective for the treatment of focal-onset seizures. This post-hoc analysis examined the effects of baseline clinical features on the efficacy of adjunctive cenobamate during the study.

METHODS

Adults with uncontrolled focal seizures despite treatment with 1-3 antiepileptic drugs/antiseizure medications (AEDs/ASMs) were randomized 1:1:1:1 to placebo or cenobamate 100, 200, or 400 mg once daily. Median percent seizure frequency reduction/28 days and ≥50% responder rates were assessed during the 12-week maintenance phase (n = 397) by number of baseline (concomitant) ASMs (1, 2, >2), median baseline seizure frequency/28 days (≤9.5 vs >9.5), and median baseline duration of epilepsy (≤23 vs >23 years).

RESULTS

For patients taking 1 concomitant ASM, median percent seizure frequency reductions ranged from 44.7% to 86.0% for cenobamate-treated patients vs 24.1% for placebo; for 2 concomitant ASMs, reductions were 41.4-57.9% with cenobamate vs 33.3% for placebo; and for >2 concomitant ASMs, reductions were 41.5-67.4% with cenobamate vs 26.4% for placebo. The highest reductions occurred in the 200- and 400-mg/day cenobamate groups. For patients with baseline seizure frequency ≤9.5, the greatest reduction in median percent seizure frequency occurred in the 200-mg/day cenobamate group (66.5%); for patients with baseline seizure frequency >9.5 the greatest reduction occurred in the 400-mg/day cenobamate group (70.7%). Similar improvements were observed when assessed by median duration of epilepsy at baseline. For cenobamate-treated patients taking 1, 2, or >2 ASMs respectively, ≥50% responder rates of up to 66.7% (400 mg), 62.2% (200 mg), and 66.0% (400 mg) were observed, vs 20.0%, 29.3%, and 23.9% for placebo, respectively; 100% seizure reductions were observed in up to 25.0% (400 mg/day), 22.2% (400 mg/day), and 19.1% (400 mg/day) of cenobamate-treated patients, vs 0%, 0%, and 2.2% for placebo, respectively. Incidence of common (≥10%) central nervous system adverse events (dizziness, somnolence, fatigue, and diplopia) were highest in the >2 ASM group, but the rates were within the range reported in the primary study.

CONCLUSIONS

Clinically relevant reductions in seizure frequency including 100% seizure reductions occurred with adjunctive cenobamate regardless of number of concomitant ASMs, baseline seizure frequency, or disease duration. The greatest reductions occurred in the 200- and 400-mg/day groups.

Comparison of topological properties of functional brain networks with graph theory in temporal lobe epilepsy with different duration of disease

Background

Our study was performed to measure the alterations in topological properties of the functional brain network of temporal lobe epilepsy (TLE) at different durations, exploring the potential progression and neuropathophysiological mechanisms of TLE.

Methods

Fifty-eight subjects, including 17 TLE patients with a disease duration of ≤5 years (TLE-SD), 20 TLE patients with a disease duration of >5 years (TLE-LD), and 21 healthy controls firstly underwent the Attention Network Test (ANT) to assess the alertness function and received the resting-state functional magnetic resonance imaging (rs-fMRI). Next, a functional brain network was set up, and then the related graph of theoretical network analysis was conducted. Finally, the correlation between network property and the neuropsychological score was analyzed.

Results

The global and local efficiencies of functional brain networks in TLE-SD patients significantly decreased and tended toward random alterations. Also, the degree centrality (DC) and nodal efficiency (Ne) in right medial pre-frontal thalamus (mPFtha) and right rostral temporal thalamus (rTtha) of TLE-SD patients significantly reduced. Further analysis showed that alertness was positively associated with the characteristic path length but negatively related to the global and local efficiencies in TLE-SD patients; alertness was negatively related to the Ne of mPFtha in TLE-LD patients.

Conclusions

Our study showed that the functional brain network of TLE patients might undergo compensatory reorganization as the disease progresses, which provides useful insights into the progression and mechanism of TLE.

Medication Beliefs and Adherence to Antiseizure Medications

Introduction

The primary objective of our study was to determine the nature of medication beliefs and their association with adherence to antiseizure medications (ASMs) among elderly epilepsy patients. Our secondary objective was to enhance the psychometric properties and factor structure parameters of the Beliefs about Medications Questionnaire (BMQ) adapted to epilepsy and affected aged subjects.

Methods

A population-based survey was performed in which older adults (≥60 years of age) were invited for a free face-to-face consultation with the specialists as well as for the collection of necessary data. The eligible subjects were those who are affected with epilepsy and having epileptic seizures of any type. In addition, the participants were required to be of any sex, currently under treatment with ASMs, resident of Tehran, and able and interested to participate independently. All were carefully examined with a reasonably detailed case-history examination. Two Persian questionnaires used were Medication Adherence Rating Scale (MARS) and BMQ. Those with a MARS score of ≥6 were considered as adherent to ASMs. All data were described in descriptive terms. We did a group comparison of means and proportions for all possible independent variables between adherents and nonadherents. Then, we did a hierarchical multiple linear regression. For this, independent variables were categorized into three different blocks: (a) sociodemographic block (Block-1), (b) treatment side-effect block (Block-2), and (c) BMQ block that included ten items of the BMQ scale (Block-3). We also did a forward step-wise linear regression by beginning with an empty model. We also estimated the psychometric properties and factor structure parameters of BMQ and its two subdomains.

Results

Of all (N = 123, mean age: 63.3 years, 74.0% males), 78.0% were adherent (mean score: 7.0, 95% CI 6.2–7.8) to ASMs. The MARS scores were not different between males and females. The mean BMQ score was 23.4 (95% CI 19.8–27.0) with the mean need score of 20.0 (95% CI 18.0–22.0) and mean concern score of 16.5 (95% CI 14.3–18.7). A positive need-concern differential was 20.4%. Upon hierarchical regression, the adjusted R² for Block-1 was 33.8%, and it was 53.8% for Block-2 and 92.2% for Block-3. Upon forward step-wise linear regression, we found that “ASMs disrupt my life” (ß −1.9, ES = −1.1, p=0.008) as the only belief associated with adherence. The alpha coefficient of BMQ was 0.81.

Conclusions

Ours is one of the very few studies that evaluated medication beliefs and their association with adherence to ASMs among elderly epilepsy patients in a non-western context. In our context, medication beliefs are likely to have an independent role in effecting adherence to ASMs, particularly the concern that “ASMs disrupt life.” Treating physicians should cultivate good conscience about ASMs and evaluate the patient's medication beliefs early-on to identify those who might be at the risk of becoming nonadherent.

Course and prognosis of adult-onset epilepsy in Brazil: A cohort study

BACKGROUND

Most of the epilepsy longitudinal studies have analyzed children. However, in endemic regions, such as Brazil, neurocysticercosis accounts for many adult-onset epilepsy cases. So, the main objective of this study was to identify the clinical predictors associated with drug-resistant adult-onset epilepsy in Brazil during a long-term follow-up.

METHODS

We followed 302 individuals with adult-onset epilepsy for 9.8 years in our University Hospital. Structured questionnaires about drug-resistant epilepsy were applied. The presence of drug-resistant epilepsy was the primary outcome. We used multilevel linear modeling in our data analysis.

RESULTS

Overall 47 (15.6%) individuals presented drug-resistant epilepsy and the etiology was structural in 70.2% of them, while infectious etiology was present in 8.5% of this group. Infectious etiology occurred in 25.9% (n = 66) of the patients from the nondrug-resistant group. Those with developmental delay were two times more likely to present seizures. Structural epilepsy etiology was associated with an increased chance of relapsing. Poor school performance and abnormal electroencephalogram were also associated with an increased chance of seizures.

CONCLUSION

The course of epilepsy was favorable in the majority of our patients, and drug-resistant epilepsy rates were similar to those found in other studies, although we evaluated older individuals with higher levels of infectious etiology. Also, we found that neurocysticercosis was associated with well-controlled epilepsy, while structural epilepsy was directly related to the occurrence of seizures. We also hypothesized that the smaller size of lesions found in neurocysticercosis could contribute to better treatment response.

Evaluation of Fear in Idiopathic Epilepsy Using Population-Based Survey and Bhalla-Gharagozli Fear in Epilepsy Questionnaire (BG-FEQ)

INTRODUCTION

The primary objective of this study was to evaluate fear related to epilepsy and its treatment among those with idiopathic epilepsy. Our secondary objective was to estimate the psychometric properties of a brief Bhalla-Gharagozli Fear in epilepsy Questionnaire (BG-FEQ).

METHODS

We conducted patient-finding exercise in our study areas through various means to obtain subjects with idiopathic epilepsy. We carefully examined each patient through a detailed case-history examination. Following that, we evaluated fear related to epilepsy by using Bhalla-Gharagozli Fear in Epilepsy Questionnaire (BG-FEQ) across two broad domains: epilepsy and pharmacotherapy.

RESULTS

The study obtained 52 subjects (39.0 years; 45.0% males, 70.0% married, 35.0% unqualified, 85.0% active epilepsy, 80.0% generalized seizures) with idiopathic epilepsy. The alpha coefficient was 92.8, with no item-specific coefficient of ≤0.91. The alpha coefficient was 0.90 and 0.93 for reporting a "yes" and "no" to the items, respectively. We obtained a two-factor structure of BG-FEQ that provided a cumulative variance of 83.6%. The majority (65.0%) reported at least one fear. The per-patient mean number of the fear element was 2.1 (95% CI 1.1-3.3), which differed significantly for males and females (1.1, 95% CI 0.4-2.6 and 3.0, 95% CI 1.4-4.6, respectively, p=0.03). The most frequent fear was that of addiction and the bad effects of anti-seizure medications (both 45.0%). Upon bootstrap regression after constraining gender, the fear elements were associated with illiteracy, difficulty in understanding epilepsy and sleeping in a prone position. The sample power was 99.0%.

CONCLUSION

There was a significant representation of fear among those with idiopathic epilepsy, especially among the females, particularly the fear of brain tumour, premature death and more frequent/severe seizures over time. At least 65.0% of idiopathic subjects are likely to be affected by at least one fear. The essential mitigating approach should be the education of practitioners towards better identification and therapeutic handling of comorbid constructs, and also for the education of patients and their caregivers towards better awareness and prevention. There is also a need for formal Epilepsy Educators towards better awareness, therapeutic support and prevention of epilepsy.

Clinical opinion: Earlier employment of polytherapy in sequential pharmacotherapy of epilepsy

Modern pharmacotherapy for epilepsy consists of orderly, sequential drug trials, in which antiepileptic drugs (AEDs) are chosen under the concept of individual patient-oriented (or - tailored) pharmacotherapy. Although monotherapy has been established as the preferred mode of AEDs therapy in both newly diagnosed and drug resistant epilepsies, there are still lack of evidence to favor either monotherapy or polytherapy in epilepsy, which has generated continuing controversies on the preferred mode of pharmacotherapy. However, each mode of pharmacotherapy may have both advantages and disadvantages, which are different and variable related to individual case scenario. We conducted a brief comparative overview between monotherapy and polytherapy to provide clues for earlier employment of polytherapy in each steps of sequential drug trials. Previous claims about the advantages of monotherapy over polytherapy are not supported but gradually losing its ground by the introduction of a large number of drugs carrying pharmacological advantages for combination therapy. Current evidence stresses the importance of combining drugs having synergistic interactions for better outcome of polytherapy, which has not been considered in previous clinical investigations comparing monotherapy and polytherapy. It is likely that a significant improvement in the outcome of current AEDs therapy is feasible by earlier employment of polytherapy as well as identification of combination drug regimens carrying synergistic interactions. At present, lamotrigine(LTG) and valproate(VPA) combination regimen is the only well documented synergistic regimen, but there are a long-list of candidate regimens requiring future trials in appropriate designs.

Prediction of antiepileptic drug treatment outcomes of patients with newly diagnosed epilepsy by machine learning

OBJECTIVE

The objective of this study was to build a supervised machine learning-based classifier, which can accurately predict the outcomes of antiepileptic drug (AED) treatment of patients with newly diagnosed epilepsy.

METHODS

We collected information from 287 patients with newly diagnosed epilepsy between 2009 and 2017 at the Second Affiliated Hospital of Zhejiang University. Patients were prospectively followed up for at least 3 years. A number of features, including demographic features, medical history, and auxiliary examinations (electroencephalogram [EEG] and magnetic resonance imaging [MRI]) are selected to distinguish patients with different remission outcomes. Seizure outcomes classified as remission and never remission. In addition, remission is further divided into early remission and late remission. Five classical machine learning algorithms, i.e., Decision Tree, Random Forest, Support Vector Machine, XGBoost, and Logistic Regression, are selected and trained by our dataset to get classification models.

RESULTS

Our study shows that 1) compared with the other four algorithms, the XGBoost algorithm based machine learning model achieves the best prediction performance of the AED treatment outcomes between remission and never remission patients with an F1 score of 0.947 and an area under the curve (AUC) value of 0.979; 2) The best discriminative factor for remission and never remission patients is higher number of seizures before treatment (>3); 3) XGBoost-based machine learning model also offers the best prediction between early remission and later remission patients, with an F1 score of 0.836 and an AUC value of 0.918; 4) multiple seizure type has the highest dependence to the categories of early and late remission patients.

SIGNIFICANCES

Our XGBoost-based machine learning classifier accurately predicts the most probable AED treatment outcome of a patient after he/she finishes all the standard examinations for the epilepsy disease. The classifier's prediction result could help disease guide counseling and eventually improve treatment strategies.

Identifying patients who will not reachieve remission after breakthrough seizures

Objective

We aim to identify people with epilepsy who are unlikely to reachieve a 12‐month remission within 2 years after experiencing a breakthrough seizure following an initial 12‐month remission.

Methods

We apply a novel longitudinal discriminant approach to data from the Standard and New Antiepileptic Drugs study to dynamically predict the risk of a patient not achieving a second remission after a breakthrough seizure by combining both baseline covariates (collected at the time of breakthrough seizure) and follow‐up data.

Results

The model classifies 83% of patients. Of these, 73% of patients (95% confidence interval [CI] = 58%‐88%) who did not achieve a second remission were correctly identified (sensitivity), and 84% of patients (95% CI = 69%‐96%) who achieved a second remission were correctly identified (specificity). The area under the curve from our model was 87% (95% CI = 80%‐94%). Patients who did not achieve a second remission were correctly identified on average after 10 months of observation postbreakthrough. Occurrence of seizures after breakthrough and the number of seizures experienced were the most informative longitudinal variables. These longitudinal profiles were influenced by the following baseline covariates: age at breakthrough seizure, presence of neurological insult, and number of antiepileptic drugs required to achieve first remission.

Significance

Using longitudinal data gathered during patient follow‐up allows more accurate predictions than using baseline covariates in a standard Cox model. The model developed in this paper is a useful first step in developing a tool for identifying patients who develop drug resistance after an initial remission.