Machine learning models for predicting seizure outcome after MR-guided laser interstitial thermal therapy in children.

OBJECTIVE: MR-guided laser interstitial thermal therapy (MRgLITT) is associated with lower seizure-free outcome but better safety profile compared to open surgery. However, the predictors of seizure freedom following MRgLITT remain uncertain. This study aimed to use machine learning to predict seizure-free outcome following MRgLITT and to identify important predictors of seizure freedom in children with drug-resistant epilepsy. METHODS: This multicenter study included children treated with MRgLITT for drug-resistant epilepsy at 13 epilepsy centers. The authors used clinical data, diagnostic investigations, and ablation features to predict seizure-free outcome at 1 year post-MRgLITT. Patients from 12 centers formed the training cohort, and patients in the remaining center formed the testing cohort. Five machine learning algorithms were developed on the training data by using 10-fold cross-validation, and model performance was measured on the testing cohort. The models were developed and tested on the complete feature set. Subsequently, 3 feature selection methods were used to identify important predictors. The authors then assessed performance of the parsimonious models based on these important variables. RESULTS: This study included 268 patients who underwent MRgLITT, of whom 44.4% had achieved seizure freedom at 1 year post-MRgLITT. A gradient-boosting machine algorithm using the complete feature set yielded the highest area under the curve (AUC) on the testing set (AUC 0.67 [95% CI 0.50-0.82], sensitivity 0.71 [95% CI 0.47-0.88], and specificity 0.66 [95% CI 0.50-0.81]). Logistic regression, random forest, support vector machine, and neural network yielded lower AUCs (0.58-0.63) compared to the gradient-boosting machine but the findings were not statistically significant (all p > 0.05). The 3 feature selection methods identified video-EEG concordance, lesion size, preoperative seizure frequency, and number of antiseizure medications as good prognostic features for predicting seizure freedom. The parsimonious models based on important features identified by univariate feature selection slightly improved model performance compared to the complete feature set. CONCLUSIONS: Understanding the predictors of seizure freedom after MRgLITT will assist with prognostication.

Number of epilepsy surgeries has decreased despite an increase in pre-surgical evaluations at a tertiary pediatric epilepsy center in Ontario.

OBJECTIVE: A recent U.S. study reported that the number of epilepsy surgeries has remained stable or declined in recent years despite an increase in pre-surgical evaluation. This study aimed to evaluate trends in pre-surgical evaluation and epilepsy surgery from 2001 to 2019 and to determine whether these trends have changed in the later period (2014-2019) compared to earlier period (2001-2013). METHODS: This study evaluated trends in pre-surgical evaluation and epilepsy surgery at a tertiary pediatric epilepsy center. Children with drug resistant epilepsy who were evaluated for surgery were included. Clinical data, reasons for not undergoing surgery, and surgical characteristics of surgery patients were collected. Overall trends and trends in later period compared to earlier period for pre-surgical evaluation and epilepsy surgery were assessed. RESULTS: There were 1151 children who were evaluated for epilepsy surgery and 546 underwent surgery. There was an upward trend in pre-surgical evaluation in the earlier period (rate ratio [RR]=1.04 (95%CI:1.02-1.07), p<0.001) and the trajectory of presurgical evaluation in the later period was not significantly different to the earlier period (RR=1.00 [95%CI:0.95-1.06], p = 0.88). Among the reasons for not undergoing surgery, failure to localize the seizures occurred more frequently in later period than earlier period (22.6% vs. 17.1% respectively, p = 0.024). For number of surgeries, there was an upward trend between 2001 and 2013 (RR=1.08 [95%CI:1.05-1.11], p<0.001), and a decreasing trend in the later period compared to earlier period (RR=0.91 [95%CI:0.84-0.99], p = 0.029). CONCLUSION: Despite an increasing trend in pre-surgical evaluation, there was a decreasing trend in the number of epilepsy surgery in the later period as there was a larger proportion of patients in whom the seizures could not be localized. Trends in presurgical evaluation and epilepsy surgery will continue to evolve with introduction of technologies such as stereo-EEG and minimally invasive laser therapy.

Seizure outcome of pediatric magnetic resonance-guided laser interstitial thermal therapy versus open surgery: A matched noninferiority cohort study.

OBJECTIVE: Minimally invasive magnetic resonance-guided laser interstitial thermal therapy (MRgLITT) has been proposed as an alternative to open epilepsy surgery, to address concerns regarding the risk of open surgery. Our primary hypothesis was that seizure freedom at 1 year after MRgLITT is noninferior to open surgery in children with drug-resistant epilepsy (DRE). The secondary hypothesis was that MRgLITT has fewer complications and shorter hospitalization than surgery. The primary objective was to compare seizure outcome of MRgLITT to open surgery in children with DRE. The secondary objective was to compare complications and length of hospitalization of the two treatments. METHODS: This retrospective multicenter cohort study included children with DRE treated with MRgLITT or open surgery with 1-year follow-up. Exclusion criteria were corpus callosotomy, neurostimulation, multilobar or hemispheric surgery, and lesion with maximal dimension > 60 mm. MRgLITT patients were propensity matched to open surgery patients. The primary outcome was seizure freedom at 1 year posttreatment. The difference in seizure freedom was compared using noninferiority test, with noninferiority margin of -10%. The secondary outcomes were complications and length of hospitalization. RESULTS: One hundred eighty-five MRgLITT patients were matched to 185 open surgery patients. Seizure freedom at 1 year follow-up was observed in 89 of 185 (48.1%) MRgLITT and 114 of 185 (61.6%) open surgery patients (difference = -13.5%, one-sided 97.5% confidence interval = -23.8% to ∞, pNoninferiority  = .79). The lower confidence interval boundary of -23.8% was below the prespecified noninferiority margin of -10%. Overall complications were lower in MRgLITT compared to open surgery (10.8% vs. 29.2%, respectively, p < .001). Hospitalization was shorter for MRgLITT than open surgery (3.1 ± 2.9 vs. 7.2 ± 6.1 days, p < .001). SIGNIFICANCE: Seizure outcome of MRgLITT at 1 year posttreatment was inferior to open surgery. However, MRgLITT has the advantage of better safety profile and shorter hospitalization. The findings will help counsel children and parents on the benefits and risks of MRgLITT and contribute to informed decision-making on treatment options.

Development and validation of machine learning models for prediction of seizure outcome after pediatric epilepsy surgery.

OBJECTIVE: There is substantial variability in reported seizure outcome following pediatric epilepsy surgery, and lack of individualized predictive tools that could evaluate the probability of seizure freedom postsurgery. The aim of this study was to develop and validate a supervised machine learning (ML) model for predicting seizure freedom after pediatric epilepsy surgery. METHODS: This is a multicenter retrospective study of children who underwent epilepsy surgery at five pediatric epilepsy centers in North America. Clinical information, diagnostic investigations, and surgical characteristics were collected, and used as features to predict seizure-free outcome 1 year after surgery. The dataset was split randomly into 80% training and 20% testing data. Thirty-five combinations of five feature sets with seven ML classifiers were assessed on the training cohort using 10-fold cross-validation for model development. The performance of the optimal combination of ML classifier and feature set was evaluated in the testing cohort, and compared with logistic regression, a classical statistical approach. RESULTS: Of the 801 patients included, 61.3% were seizure-free 1 year postsurgery. During model development, the best combination was XGBoost ML algorithm with five features from the univariate feature set, including number of antiseizure medications, magnetic resonance imaging lesion, age at seizure onset, video-electroencephalography concordance, and surgery type, with a mean area under the curve (AUC) of .73 (95% confidence interval [CI] = .69-.77). The combination of XGBoost and univariate feature set was then evaluated on the testing cohort and achieved an AUC of .74 (95% CI = .66-.82; sensitivity = .87, 95% CI = .81-.94; specificity = .58, 95% CI = .47-.71). The XGBoost model outperformed the logistic regression model (AUC = .72, 95% CI = .63-.80; sensitivity = .72, 95% CI = .63-.82; specificity = .66, 95% CI = .53-.77) in the testing cohort (p = .005). SIGNIFICANCE: This study identified important features and validated an ML algorithm, XGBoost, for predicting the probability of seizure freedom after pediatric epilepsy surgery. Improved prognostication of epilepsy surgery is critical for presurgical counseling and will inform treatment decisions.

Health Care System Costs Associated With Surgery and Medical Therapy for Children With Drug-Resistant Epilepsy in Ontario.

BACKGROUND AND OBJECTIVES: Improvement in seizure control after epilepsy surgery could lead to lower health care resource use and costs, but it is uncertain whether this could offset the high costs related to surgery. This study aimed to evaluate phase-specific and cumulative long-term health care costs of surgery compared to medical therapy in children with drug-resistant epilepsy from the health care payer perspective. METHODS: Children who were evaluated for epilepsy surgery and treated with surgery or medical therapy from 2003 to 2018 at the Hospital for Sick Children in Toronto were identified from chart review and linked to their health administrative databases in Ontario, Canada. Inverse probability of treatment weighting with stabilized weights was used to balance the baseline covariates between the 2 groups. Patients were assigned to presurgery, surgery, short-term (first 2 years), intermediate-term (2-5 years), and long-term (>5 years) postsurgery care phases on the basis of treatment trajectory. Phase-specific and cumulative long-term health care costs were evaluated. Costs were converted from Canadian to US dollars year 2018 value. RESULTS: There were 372 surgical and 258 medical patients. Costs were higher in surgical than medical patients for presurgery (3 and 39 weeks), surgery, and short-term care phase, and the attributable costs of surgery per 7 patient-days were $1,602 (95% CI $1,438-$1,785), $172 (95% CI $147-$185), $19,819 (95% CI $18,822-$20,932), and $28 (95% CI $22-$32), respectively. Costs were lower in surgical patients for intermediate- and long-term care phase, and the attributable costs were -$72 (95% CI -$124 to -$35) and -$94 (95% CI -$129 to -$63), respectively. In surgical patients, costs were highest for surgery followed by presurgery care phase, with hospitalizations accounting for the highest cost component. In medical patients, costs increased gradually from presurgery to long-term care phase. Cumulative costs were higher for surgical than medical patients in the first 7 years after surgery, but from 8 years on, costs were lower for surgical patients. DISCUSSION: This study demonstrated the long-term economic benefits of epilepsy surgery compared to medical therapy for the health care system with the use of real-world data, which would justify the high costs of surgery. The results will support future economic evaluation comparing minimally invasive treatment such as laser therapy to surgery.

Vigabatrin-related adverse events for the treatment of epileptic spasms: systematic review and meta-analysis.

Objective: Although vigabatrin (VGB) is effective and well tolerated for the treatment of epileptic spasms, there are safety concerns. The aim of this systematic review and metaanalysis was to assess adverse events of VGB for the treatment of epileptic spasms. Methods: MEDLINE, EMBASE, and Cochrane databases were searched. The population was infants treated with VGB for epileptic spasms. The outcomes were VGB-related adverse events. Meta-analyses of VGB-related MRI abnormalities, retinal toxicity as measured by electroretinogram (ERG), visual field defect as measured by perimetry, and other adverse events were conducted. Results: Fifty-seven articles were included in the systematic review. The rate of VGB-related MRI abnormalities was 21% (95% CI: 15-29%). Risk factors for MRI abnormalities were age younger than 12 months and higher VGB dose. VGB-related retinal toxicity and visual field defect occurred in 29% (95% CI: 7-69%) and 28% (95% CI: 4-78%) respectively. Other adverse events occurred in 23% (95% CI: 16-34%), consisting predominantly of central nervous system symptoms, and the majority of these did not require therapeutic modification. Conclusion: This study will inform physicians and families on the risk profile of VGB for the treatment of epileptic spasms and will help decisions on treatment options.