Presurgical differentiation between malignant haemangiopericytoma and angiomatous meningioma by a radiomics approach based on texture analysis

PurposeTo assess whether a machine-learning model based on texture analysis (TA) could yield a more accurate diagnosis in differentiating malignant haemangiopericytoma (HPC) from angiomatous meningioma (AM).Materials and methodsSixty-seven pathologically confirmed cases, including 24 malignant HPCs and 43 AMs between May 2013 and September 2017 were retrospectively reviewed. In each case, 498 radiomic features, including 12 clinical features and 486 texture features from MRI sequences (T2-FLAIR, DWI and enhanced T1WI), were extracted. Three neuroradiologists independently made diagnoses by vision. Four Support Vector Machine (SVM) classifiers were built, one based on clinical features and three based on texture features from three MRI sequences after feature selection. The diagnostic abilities of these classifiers and three neuroradiologists were evaluated by receiver operating characteristic (ROC) analysis.ResultsMalignant HPCs were found to have larger sizes, slighter degrees of peritumoural oedema compared with AMs (P < 0.05), and more serpentine-like vessels. The AUC of the enhanced T1WI-based classifier was 0.90, significantly higher than that of T2-FLAIR-based or DWI-based classifiers (0.77 and 0.73). The AUC of the SVM classifier based on clinical features was 0.66, slightly but not significantly lower than the performances of 3 neuroradiologists (AUC = 0.69, 0.70 and 0.73).ConclusionMachine-learning models based on clinical features alone could not provide a better diagnostic performance than that of radiologists. The SVM classifier built by texture features extracted from enhanced T1WI is a promising tool to differentiate malignant HPC from AM before surgery.     

Treatment and survival of supratentorial and posterior fossa ependymomas in adults

Ependymoma is a rare primary brain or spinal cord tumor that arises from the ependyma, a tissue of the central nervous system. This study analyzed a large cohort of adult supratentorial and posterior fossa ependymoma tumors in order to elucidate factors associated with overall survival. We utilized the USA National Cancer Database to study adult World Health Organization grade II/III supratentorial and posterior fossa ependymoma patients treated between 1998 and 2011. Overall survival was estimated by the Kaplan–Meier method and factors associated with survival were determined using a multivariate Cox proportional hazards model. Among 1318 patients, 1055 (80.0%) had grade II and 263 (20.0%) anaplastic tumors located in the posterior fossa (64.3%) and supratentorial region (35.7%). Overall average age was 44.3 years, 48.0% of patients were female, 86.5% were Caucasian, and 36.8% underwent near/gross total surgical resection. Radiotherapy was given to 662 patients (50.8%) and 75 (5.9%) received chemotherapy. Older age at diagnosis (hazard ratio [HR] 1.51, p < 0.0001), high tumor grade (HR 1.82, p = 0.005), and large tumor size (HR 1.66, p = 0.008) were associated with poor survival. Females compared to males (HR 0.67, p = 0.03) and patients with posterior fossa tumors versus supratentorial (HR 0.64, p = 0.04) had a survival advantage. Our study showed that older patients, with supratentorial tumors, and high histological grade had an increased risk of mortality. A survival benefit was captured in females and patients with posterior fossa tumors. Adjuvant radiotherapy and chemotherapy did not confer a survival benefit among all patients, even after stratification by tumor grade or anatomical location.     

Band-sensitive seizure onset detection via CSP-enhanced EEG features

This paper presents two novel epileptic seizure onset detectors. The detectors rely on a common spatial pattern (CSP)-based feature enhancement stage that increases the variance between seizure and nonseizure scalp electroencephalography (EEG). The proposed feature enhancement stage enables better discrimination between seizure and nonseizure features. The first detector adopts a conventional classification stage using a support vector machine (SVM) that feeds the energy features extracted from different subbands to an SVM for seizure onset detection. The second detector uses logical operators to pool SVM seizure onset detections made independently across different EEG spectral bands. The proposed detectors exhibit an improved performance, with respect to sensitivity and detection latency, compared with the state-of-the-art detectors. Experimental results have demonstrated that the first detector achieves a sensitivity of 95.2%, detection latency of 6.43 s, and false alarm rate of 0.59 per hour. The second detector achieves a sensitivity of 100%, detection latency of 7.28 s, and false alarm rate of 1.2 per hour for the MAJORITY fusion method.     

Ectopic recurrence of craniopharyngioma in the posterior fossa: Case report and review of the literature

Craniopharyngiomas are benign epithelial tumors which may very occasionally recur in ectopic locations. We present two cases of ectopic recurrence, both in the posterior fossa, and provide a review of the literature with basic statistics. Two patients admitted to our institution with posterior fossa lesions underwent gross total resection. Pathological studies showed adamantinomatous craniopharyngiomas (ACP). Both patients had a prior history of suprasellar tumor surgery. We also reviewed the related data and undertook a basic statistical analysis. We found 67 cases of ectopic recurrent craniopharyngioma (including the present cases): 51 cases were adamantinomatous (76%), 6 papillary (9%) and 10 unknown (15%). 18 cases occurred in the posterior fossa, all of them diagnosed as the ACP subtype. The intervals until recurrence were 15.15 years for posterior fossa recurrences and 5.75 years for supratentorial cases. Student t test showed significant differences in time to recurrence (p 0.002). Gross total resection was performed in 53 cases (79%), subtotal resection + radiotherapy in 3 (5%) and 11 (16%) cases were treated with other options. Ectopic recurrence is a rare but possible event. Those in the posterior fossa may appear later than in the supratentorial space. ACP is likely to be the most common subtype in these cases, possibly due to its more aggressive behavior compared to the papillary subtype. Long term follow-up should be performed to detect ectopic recurrences. Ectopic recurrences are often surgically accessible and gross total resection should be achieved.     

Predicting outcome in childhood diffuse midline gliomas using magnetic resonance imaging based texture analysis

BackgroundDiffuse midline gliomas (DMG) are aggressive brain tumours, previously known as diffuse intrinsic pontine gliomas (DIPG), with 10% overall survival (OS) at 18 months. Predicting OS will help refine treatment strategy in this patient group. MRI based texture analysis (MRTA) is novel image analysis technique that provides objective information about spatial arrangement of MRI signal intensity (heterogeneity) and has potential to be imaging biomarker.ObjectivesTo investigate MRTA in predicting OS in childhood DMG.MethodsRetrospective study of patients diagnosed with DMG, based on radiological features, treated at our institution 2007–2017. MRIs were acquired at diagnosis and 6 weeks after radiotherapy (54 Gy in 30 fractions). MRTA was performed using commercial available TexRAD research software on T2 W sequence and Apparent Diffusion Coefficient (ADC) maps encapsulating tumour in the largest single axial plane. MRTA comprised filtration-histogram technique using statistical and histogram metrics for quantification of texture. Kaplan-Meier survival analysis determined association of MRI texture parameters with OS.ResultsIn all, 32 children 2–14 years (median 7 years) were included. MRTA was undertaken on T2W (n = 32) and ADC (n = 22). T2W-MRTA parameters were better at prognosticating than ADC-MRTA. Children with homogenous tumour texture, at medium scale on diagnostic T2W MRI, had worse prognosis (Mean of Positive Pixels (MPP): P = 0.005, mean: P = 0.009, SD: P = 0.011, kurtosis: P = 0.037, entropy: P = 0.042). Best predictor MPP was able to stratify patients into poor and good prognostic groups with median survival of 7.5 months versus 17.5 months, respectively.ConclusionsDMG with more homogeneous texture on diagnostic MRI is associated with worse prognosis. Texture parameter MPP is the most predictive marker of OS in childhood DMG.     

Growth rate of vestibular schwannoma

Vestibular schwannoma (VS) is the most common tumor in the extra-axial posterior fossa compartment in adults. Growth rate is paramount to decision making regarding treatment and follow up of these tumors. We conducted a comprehensive review of the literature to answer four questions: What percentage of newly diagnosed VS will grow on follow-up? What factors correlate to tumor growth? What is the “normal” growth rate for sporadic VS? What factors characterize VS with rapid growth? Thirty-seven reports, with more than 4000 patients, fit our review criteria. One third of newly diagnosed VS will grow on follow-up of 1–3 years. However, after 5 years, up to one half will grow. Patient age and sex do not influence growth of VS. Hearing loss and vertigo at presentation do not predict tumor growth. It is unclear whether balance disturbance or tinnitus predict tumor growth. Tumor size and location do not predict tumor growth. Growth in the first year of observation is a strong predictor of tumor growth. The average growth rate of a VS is 0.99–1.11 mm/year. However, the expected growth rate for VS that have been shown to grow at first follow-up is 3 mm/year. Factors that may predict tumor growth of above 4 mm/year are cystic and hemorrhagic features in the tumor, and hormonal treatment. VS grow at an average 1 mm/year. VS that have been shown to grow at first follow-up should be considered for treatment, unless contraindicated. Long term follow-up is recommended for VS.     

Distinguishing age-related cognitive decline from dementias: A study based on machine learning algorithms

Background and aimThis study aims to examine the distinguishability of age-related cognitive decline (ARCD) from dementias based on some neurocognitive tests using machine learning.Materials and methods106 subjects were divided into four groups: ARCD (n = 30), probable Alzheimer’s disease (AD) (n = 20), vascular dementia (VD) (n = 21) and amnestic mild cognitive impairment (MCI) (n = 35). The following tests were applied to all subjects: The Wechsler memory scale-revised, a clock-drawing, the dual similarities, interpretation of proverbs, word fluency, the Stroop, the Boston naming (BNT), the Benton face recognition, a copying-drawings and Öktem verbal memory processes (Ö-VMPT) tests. A multilayer perceptron, a support vector machine and a classification via regression with M5-model trees were employed for classification.ResultsThe pairwise classification results show that ARCD is completely separable from AD with a success rate of 100% and highly separable from MCI and VD with success rates of 95.4% and 86.30%, respectively. The neurocognitive tests with the higher merit values were Ö-VMPT recognition (ARCD vs. AD), Ö-VMPT total learning (ARCD vs. MCI) and semantic fluency, proverbs, Stroop interference and naming BNT (ARCD vs. VD).ConclusionThe findings show that machine learning can be successfully utilized for distinguishing ARCD from dementias based on neurocognitive tests.     

Traumatic interhemispheric subdural haematoma: Study of 35 cases

The clinical and radiological findings, management, and outcomes in 35 patients with traumatic interhemispheric subdural haematoma (ISH) were reviewed retrospectively. Twenty-five patients had favourable outcomes and 10 had poor outcomes. All patients were treated conservatively for ISH. Univariate analysis found that the Glasgow Coma Scale (GCS) score (p < 0.001), hypovolemic shock (p = 0.018), skull fracture (p = 0.008), convexity or posterior fossa subdural haematoma (p = 0.008), and subarachnoid haemorrhage (SAH) were correlated with outcome (p < 0.001). Multivariate analysis showed that GCS score (p = 0.031; odds ratio [OR], 0.6; 95% confidence interval [CI], 0.3–0.9) and the presence of SAH (p = 0.023; OR, 14.2; 95% CI, 1.5–138.2) were significantly related to poor outcome. This study provides important information on the clinicoradiological findings and prognoses in patients with traumatic ISH.     

Outcomes of middle fossa craniotomy for the repair of superior semicircular canal dehiscence

Superior semicircular canal dehiscence (SSCD) is a rare defect of the arcuate eminence that causes an abnormal connection between the superior semicircular canal and middle cranial fossa. Patients often present with a variety of auditory and vestibular symptoms. Trigger avoidance is the initial strategy, but surgery may be necessary in debilitating cases. We retrospectively reviewed SSCD patients undergoing repair via a middle fossa craniotomy between March 2011 and September 2015. Forty-nine patients undergoing 58 surgeries were identified. Autophony was the most common symptom at presentation (n = 44; 90%). Mean follow-up was 10.9 months, with 100% of patients reporting resolution of at least one symptom. Aural fullness was the most commonly resolved symptom following surgical repair (n = 19/22; 86%). Hearing loss (n = 11/25; 44%) and tinnitus (n = 11/38; 29%) were the most common symptoms to persist following surgery. The most common symptom to develop after surgery was disequilibrium (n = 4/18; 22%). Upon comparing the overall pre-operative and post-operative groups, the number of patients with autophony (p < 0.0001), aural fullness (p = 0.0006), hearing loss (p = 0.0119), disequilibrium (p = 0.0002), sound- and pressure-induced vertigo (p < 0.0001), and tinnitus (p < 0.0001) were significantly different. Improved clinical outcomes were demonstrated in patients undergoing SSCD repair through a middle cranial fossa approach. The most common presenting symptom (autophony) was also most likely to resolve after surgery. Hearing loss is less amenable to surgical correction. Disequilibrium developed in a small number of patients after repair.     

Transorbital neuroendoscopic surgery for the treatment of skull base lesions

Transorbital neuroendoscopic surgery (TONES) is a relatively new technique that not only allows access to the contents of the orbit but also the intracranial compartment, including the anterior cranial fossa, middle fossa and lateral cavernous sinus. In this study, we aimed to retrospectively review the largest experience to our knowledge with regards to surgical outcomes of skull base pathologies treated with a TONES procedure. Forty patients (aged 3–89 years) underwent 45 TONES procedures between the years of 2006–2013. Pathologies were cerebrospinal fluid leak repair (n = 16), traumatic fracture (n = 8), tumor (n = 11), meningoencephalocele (n = 5), hematoma (n = 1), and infection (n = 4). Three patients had a persistent complication at 3 months, including a case each of enophthalmos (unnoticed by patient), epiphora (delayed presentation at 2 months requiring dacryocystorhinostomy), and ptosis (improved at 1 year). Surgical success was achieved in all patients. Of special import, there were no cases of visual decline, diplopia, or stroke. There was no mortality. To our knowledge this is the first study and largest experience of TONES (level 4 evidence) to detail outcomes with respect to skull base pathologies. Our results indicate that TONES procedures can be performed with minimal morbidity. Further studies are needed to assess equivalency with craniotomy based approaches though this initial report is encouraging.     

Results of phase II pharmacokinetic study of levetiracetam for prevention of post‐traumatic epilepsy

Levetiracetam (LEV) has antiepileptogenic effects in animals and is a candidate for prevention of epilepsy after traumatic brain injury. Pharmacokinetics of LEV in TBI patients was unknown. We report pharmacokinetics of TBI subjects ≥ 6 years with high PTE risk treated with LEV 55 mg/kg/day orally, nasogastrically or intravenously for 30 days starting ≤ 8 h after injury in a phase II safety and pharmacokinetic study. Forty‐one subjects (26 adults and 15 children) were randomized to PK studies on treatment days 3 and 30. Thirty‐six out of forty‐one randomized subjects underwent PK study on treatment day 3, and 24/41 subjects underwent PK study on day 30. On day 3, mean T_max was 2.2 h, C_max was 60.2 μg/ml and AUC was 403.7 μg/h/ml. T_max was longer in the elderly than in children and non-elderly adults (5.96 h vs. 1.5 h and 1.8 h; p = 0.0001). AUC was non-significantly lower in children compared with adults and the elderly (317.4 μg/h/ml vs. 461.4 μg/h/ml and 450.2 μg/h/ml; p = 0.08). C_max trended higher in i.v.- versus tablet- or n.g.-treated subjects (78.4 μg/ml vs. 59 μg/ml and 48.2 μg/ml; p = 0.07). AUC of n.g. and i.v. administrations was 79% and 88% of AUC of oral administration. There were no significant PK differences between days 3 and 30. Treatment of TBI patients with high PTE risk with 55 mg/kg/day LEV, a dose with antiepileptogenic effect in animals, results in plasma LEV levels comparable to those in animal studies.     

Implicit and explicit motor learning: Application to children with Autism Spectrum Disorder (ASD)

Aims and objectivesThis study aims to determine whether children with Autism Spectrum Disorder (ASD) are capable of learning a motor skill both implicitly and explicitly.MethodsIn the present study, 30 boys with ASD, aged 7–11 with IQ average of 81.2, were compared with 32 typical IQ- and age-matched boys on their performance on a serial reaction time task (SRTT). Children were grouped by ASD and typical children and by implicit and explicit learning groups for the SRTT.ResultsImplicit motor learning occurred in both children with ASD (p = .02) and typical children (p = .01). There were no significant differences between groups (p = .39). However, explicit motor learning was only observed in typical children (p = .01) not children with ASD (p = .40). There was a significant difference between groups for explicit learning (p = .01).DiscussionThe results of our study showed that implicit motor learning is not affected in children with ASD. Implications for implicit and explicit learning are applied to the CO-OP approach of motor learning with children with ASD.     

Hemangiopericytoma: Radical resection remains the cornerstone of therapy

Hemangiopericytomas (HPC) are mesenchymal tumors with a propensity towards chronicity and metastasis. This study aimed to reflect a single institution experience with both World Health Organization (WHO) grade II and III HPC. Pathology records from the years 1990–2013 at the University of Washington were searched to identify tumors unequivocally classified as HPC. Electronic chart review was then utilized to collect pertinent patient data. Of the WHO grade II HPC, there were four men and two women (average age 52 years) while the grade III HPC group had eight men and two women (average age 51 years). Sixty-six percent of WHO grade II tumors were located in the middle or posterior fossa as compared to none of the grade III tumors. Survival analysis revealed a significant survival benefit for patients who underwent complete resection (223 months) versus those with subtotal resection (138 months, p < 0.05). Factors such as age, sex, the use of up-front radiation, and whether the patient had a recurrence did not show statistical significance related to overall survival or progression free survival. Radiation in the form of external beam radiotherapy given at the time of the first recurrence did trend towards improved progression free survival (56 months) compared to those patients who were not radiated (22 months, p = 0.09) All patients with radical resection went on to never have a recurrence. Our results indicate that HPC are tumors with limited response to radiation and best treated with aggressive resection. Future studies will determine whether molecular-based therapies may provide added adjuvant benefit.     

Multifocal and multicentric glioblastoma: Improved characterisation with FLAIR imaging and prognostic implications

Glioblastoma usually presents on imaging as a single peripherally enhancing lesion, but multiple enhancing lesions can occur, termed multifocal if there is a connection between enhancing lesions, or multicentric when no communication is demonstrated. We aim to determine the incidence and prognostic implications of multifocal and multicentric glioblastoma in the era of modern MRI, focusing on the added benefit of T2-weighted fluid-attenuated inversion recovery (FLAIR) imaging. Patients with a new diagnosis of glioblastoma were identified. Preoperative MRI were reviewed to determine whether more than one distinct enhancing lesion was present, and whether there was communication between lesions. The findings were compared against survival data. More than one discrete contrast-enhancing lesion was present in 51 of the 151 patients (34%). Communication between lesions was identified in 47 of these, most commonly direct parenchymal spread (41 patients). The patients with multiple lesions had worse survival (median 176 days, compared to 346 days), but this difference was not statistically significant (p = 0.253). These tumours more frequently involved deep structures (p < 0.001) and the posterior fossa (p = 0.045), both of which were associated with worse survival. The presence of multiple enhancing foci in glioblastoma is common, occurring in about one-third of patients, and the majority have multifocal disease. The FLAIR sequence is the crucial sequence for demonstrating a communication between lesions. The worse survival of these patients is, at least in large part related to more extensive tumour dissemination and more frequent involvement of key structures, rather than multiplicity per se.     

Relationship between brain interstitial fluid tumor necrosis factor-α and cerebral vasospasm after aneurysmal subarachnoid hemorrhage

Tumor necrosis factor-α (TNF-α) has a crucial role in the onset of hemolysis-induced vascular injury and cerebral vasoconstriction. We hypothesized that TNF-α measured from brain interstitial fluid would correlate with the severity of vasospasm following aneurysmal subarachnoid hemorrhage (aSAH). From a consecutive series of 10 aSAH patients who underwent cerebral microdialysis (MD) and evaluation of vasospasm by CT angiogram (CTA) or digital subtraction angiography (DSA), TNF-α levels from MD were measured at 8-hour intervals from aSAH days 4–6 using enzyme-linked immunosorbent assay. An attending neuroradiologist blinded to the study independently evaluated each CTA and DSA and assigned a vasospasm index (VI). Five patients had a VI < 2 and 5 patients had a VI > 2, where the median VI was 2 (range 0–13). The median log TNF-α area under the curve (AUC) was 1.64 pg/mL 1 day (interquartile range 1.48–1.71) for the VI < 2 group, and 2.11 pg/mL 1 day (interquartile range 1.95–2.47) for the VI > 2 group (p < 0.01). Thus, in this small series of poor-grade aSAH patients, the AUC of TNF-α levels from aSAH days 4–6 correlates with the severity of radiographic vasospasm. Further analysis in a larger population is warranted based on our preliminary findings.     

CSF neurofilament light chain is elevated in OMS (decreasing with immunotherapy) and other pediatric neuroinflammatory disorders

Using a panel of seven brain cell-specific biomarkers in cerebrospinal fluid (CSF), pediatric opsoclonus–myoclonus syndrome (OMS) (n = 234) was compared to pediatric non-inflammatory neurological controls (n = 84) and other inflammatory neurological disorders (OIND) (n = 44). Only CSF NFL was elevated in untreated OMS versus controls (+ 83%). It was 87% higher in OIND than in OMS. On combination treatment with front-loaded ACTH, IVIg, rituximab, median CSF NFL decreased by 60% to control levels. These biochemical data suggest neuronal/axonal injury in some children with OMS without indicators of astrogliosis, and reduction on sufficient immunotherapy.     

The predictive validity of common risk assessment tools in men with intellectual disabilities and problematic sexual behaviors

This CIHR-funded study examined whether certain current risk assessment tools were effective in appraising risk of recidivism in a sample of sex offenders with intellectual disabilities (ID). Fifty men with ID who had engaged in problematic sexual behavior (PSB) were followed for an average of 2.5 years. Recidivism was defined and measured as any illegal or problematic behavior, as well as any problematic but not necessarily illegal behavior. At the beginning of the study, each participant was rated on two risk assessment tools: the Violence Risk Appraisal Guide (VRAG) and the Sex Offender Risk Appraisal Guide (SORAG). During each month of follow-up, participants were also rated on the Short-Dynamic Risk Scale (SDRS), an assessment tool intended to measure the risk of future problematic behaviors. Data was analyzed using t-tests, Cohen’s d and area under the curve (AUC) to test predictive validity of the assessment tools. Using the AUC, results showed that the VRAG was predictive of sexual (AUC = 0.74), sexual and/or violent (AUC = 0.71) and of any criminally chargeable event (AUC = 0.69). The SORAG was only significantly predictive of sexual events (AUC = 0.70) and the SDRS was predictive of violent events (AUC = 0.71). The t-test and Cohen’s d analyses, which are less robust to deviations from the assumptions of normal and continuous distribution than AUC, did not yield significant results in each category, and therefore, while the results of this study suggest that the VRAG and the SORAG may be effective tools in measuring the short term risk of sexual recidivism; and the VRAG and SDRS may be effective tools in appraising long term risk of sexual and/or violent recidivism in this population, it should be used with caution. Regardless of the assessment tool used, risk assessments should take into account the differences between sex offenders with and without ID to ensure effective measurement.     

Operative fusion of multilevel cervical spondylotic myelopathy: Impact of patient demographics

Cervical spondylotic myelopathy (CSM) is the most common cause of spinal cord dysfunction in patients older than age 55, with operative management being a widely adopted approach. Previous work has shown that private insurance status, gender and patient race are predictive of the operative approach patients receive (anterior-only, posterior-only, combined anterior–posterior). The Nationwide Inpatient Sample from 2001 to 2010 was used to assess the potential role of multilevel CSM as a contributing factor in determining which operative approach CSM patients receive, as it is rare for an anterior-only approach to be sufficient for CSM patients requiring fusion of four or more involved levels. Multivariate analyses revealed that female sex (OR = 3.78; 95% CI = 2.08–6.89; p < 0.0001), private insurance (OR = 5.02; 95% CI = 2.26–11.12; p < 0.0001), and elective admission type (OR = 4.12; 95% CI = 1.65–10.32; p = 0.0025) were predictive of increased receipt of a 3+ level fusion in CSM. No other variables, including patient age, race, income, or admission source were predictive of either increased or decreased likelihood of receiving fusion of at least three levels for CSM. In conclusion, female sex, private insurance status, and elective admission type are each independent predictors in CSM for receipt of a 3+ level fusion, while patient age, race and income are not. Given the propensity of fusions greater than three levels to require posterior approaches and the association between posterior CSM approaches and increased morbidity/mortality, these findings may prove useful as to which patient demographics are predictive of increased morbidity and mortality in operative treatment of CSM.     

What do epileptologists recommend about discontinuing antiepileptic drugs for a second time in children?

ObjectiveThere is a broad consensus that antiepileptic drugs (AEDs) may be withdrawn after two years of seizure freedom for most children with epilepsy. If seizures recur and are, again, completely controlled with AEDs, little is known about discontinuing a second time. We surveyed American and Canadian pediatric epileptologists to understand their current practice.MethodsIn 2014, a survey was sent via e-mail to 193 pediatric epileptologists to learn about AED discontinuation practices in children. The survey asked direct questions about practice and posed five “real-life” cases where the decision to discontinue might be difficult. Participants were identified through membership lists of several US and Canadian epilepsy organizations.ResultsThere were 94 (49%) completed surveys. Sixty-three participants had ≥ 10 years in practice (“more experienced”: mean 23 ± 9 years), and 31 had < 10 years (“less experienced”: mean 6 ± 2). Overall, 62% recommended AED discontinuation for the first time after 2–3 years of seizure freedom, and 61% recommended discontinuation for the second time after 2–3 years. Fifty-six percent of “more experienced” clinicians required a longer seizure-free period prior to a second discontinuation (p < 0.001) compared with 26% of “less experienced” clinicians (p = ns). Overall, most participants suggested an AED taper duration of 2–6 months for the first and second attempts, 52% and 68%, respectively. Both groups wean AEDs more slowly during the second attempt (p < 0.001). There was only 40–60% agreement among participants to discontinue AEDs in four of the cases.ConclusionNearly half (46%) of pediatric epileptologists require a longer seizure-free period the second time they attempt to discontinue AEDs compared with the first attempt and wean down AEDs somewhat more slowly. Although a variety of factors influence decision-making, there was a high level of disagreement to discontinue AEDs a second time in “real-life” cases.