Seizure Outcome and Its Prognostic Predictors After Hemispherotomy in Children With Refractory Epilepsy in a Chinese Pediatric Epileptic Center

Object: To explore the post-hemispherotomy seizure outcome and its prognostic predictors in children with refractory epilepsy. Methods: We reviewed 83 consecutive child patients with refractory epilepsy who underwent a hemispherectomy from June 2014 to January 2017 at our Pediatric Epilepsy Center. Demographic, clinical, EEG, neuroimaging, and surgical data were collected. Seizure outcome data were collected via outpatient clinics as well as telephone visits and were graded according to Engel criteria. Logistic regression model and Cox proportional hazard regression model were, respectively, applied to explore the related factors predicting the seizure outcomes of children after a hemispherotomy. Results: Of the 83 patients, 55 (63.2%) were male. The mean seizure onset age was 1.9 years (0–8.7 years), and the mean surgery age was 5 years (0.8–14 years). At a mean follow-up of 3 years, 69 children (83.1%) were seizure free, and 14 (16.9%) exhibited seizure recurrence. In a univariate analysis, whether or not considering follow-up time, a non-lateralized interictal EEG pattern, bilateral PET abnormalities and acute postoperative seizures (APOS) could all predict poor seizure outcomes post-hemispherotomy. Bilateral PET abnormalities were independently correlated with unfavorable seizure outcomes in the multivariate Logistic regression analysis (Odds ratio(OR) = 13.05, 95%CI = 1.52–112.29, P = 0.019) and in the multivariate Cox proportional hazard analysis(OR = 13.99, 95%CI = 2.75–71.17, P = 0.001). Conclusions: Child epileptic patients with bilateral PET abnormalities may have poor seizure outcomes after a hemispherotomy procedure. This study will facilitate better candidate selection for hemispherotomies and early identification of unfavorable seizure outcomes.


INTRODUCTION
Hemispherectomy is an effective surgical approach for medically intractable epilepsy in children with multilobar or hemispheric epileptogenic lesions, such as encephalomalacia, Rasmussen encephalitis, and malformation of cortical development (1,2). In published large-sample studies, the rate of seizure remission after hemispherectomy varies from 65 to 80% (3)(4)(5)(6)(7).Many researchers attempted to identify prognostic predictors for postoperative seizure outcomes. However, inconsistent predictors have been reported in different studies. Predictors such as the etiology of epilepsy (6), contralateral MRI findings (4,8), electroencephalography (EEG) findings (7), and surgical techniques (9) have been primarily discussed. However, other potential factors that may affect seizure outcomes including seizure semiology (7) and acute postoperative seizures (APOS) (7) received little attention. The goals of our study were to investigate seizure outcomes in a cohort of Chinese children who underwent peri-insular hemispherotomies, and to explore the potential predictors of the seizure outcomes.

Patients
We retrospectively reviewed the data of a cohort of consecutive children who underwent peri-insular hemispherotomies for medically refractory epilepsy in the Pediatric Epilepsy Center of Peking University First Hospital between June 2014 and January 2017. Eighty-three children who were followed up at least 2 years after surgery were enrolled. This study was approved by the institutional review board of Ethic Committee in Peking University First Hospital and a written informed consent was given by all the participants.

Etiology
The etiology of epilepsy was classified as three types according to neuroimaging and histopathology results: (1) developmental form (hemimegalencephaly or other malformations of cortical development), (2) acquired form (stable and non-progressive encephalomalacia, such as ischaemic strokes, prior intracerebral hemorrhage, asymmetric hypoxic ischaemic injury, prior head trauma, and so on), and (3) progressive form (Rasmussen encephalitis and Sturge Weber Syndrome).

Basic Characteristics
Demographics, age of seizure onset, age of surgery, course of epilepsy, seizure frequency, and the number of preoperative antiepileptic medicine were recorded. The developmental status before seizure onset was dichotomous as normal or delayed. Seizure frequency was determined according to parental and EEG reports.

Preoperative EEG and Semiology
Interictal/ictal scalp EEGs were recorded using a video-EEG(vEEG) monitoring system (Nihon Kohden; Japan); the electrodes were arranged according to the international 10-20 system. The duration of vEEG monitoring ranged from 1 to 7 days, and at least three habitual seizures were recorded for each patient. We attempted to document every seizure type if the patient had more than one type of seizure. Seizure types were classified by two epileptologists according to the vEEG findings.

Preoperative Neuroimaging
All patients (n = 83) underwent 3T high-resolution brain MRI (Philips Achieva, Dutch) dedicated to seizure focus identification. Subtle contralateral abnormalities such as small single white matter lesions, myelination anomalies or isolated areas of subtle T2 hyperintensity without ventricular dilatation, tissue loss, and cortical/subcortical involvement were not further analyzed. Fifty-one patients underwent an interictal FDG-PET (Philips Gemini GXL, Dutch) scan. MRI and FDG-PET results were categorized as follows: (1) unilateral (multilobar structural or metabolic abnormalities limited to one hemisphere) and (2) bilateral (structural or metabolic abnormalities observed on the contralateral hemisphere).

Peri-Insular Hemispherotomy Surgery
All resections were performed by the same chief neurosurgeon (Lx Cai) using the peri-insular hemispherotomy technique. The surgical side, postoperative complications, and the presence of APOS were recorded. APOS were defined as seizures within 7 days after surgery confirmed by vEEG.

Follow-Up
The protocol for postoperative follow-up in our epilepsy center refers to visits at 3 and 6 months followed by yearly visits after the operations, while more frequent follow-up visits are required for patients with recurrent seizures. Postoperative seizure status was assessed using a structured questionnaire and was confirmed against medical records via telephone and outpatient visits. The late remission was defined as seizure-recurrence after surgery, but absence of any seizures for at least more than 2 year at last followup. Seizure outcomes were described according to the Engel criteria (10), and the postoperative number of antiepileptic drugs was documented from the last medical records during follow-up.

Statistical Analysis
Categorical variables were summarized as counts and percentages of the total number of patients in each category. Continuous variables were expressed as the means with standard deviations or medians with ranges. Regardless of follow-up time, the initial comparisons between the seizure-free subgroup with seizure-recurrence subgroup were performed univariably using the Wilcoxon rank sum, χ 2 or Fisher's exact tests. Variables with a significance level above 10% in the initial univariate analysis were then tested in a multivariate Logistic regression model. In consideration of follow-up time, Kaplan-Meier survival analysis was performed to reveal the probability of seizure-free status in the overall group and to explore the univariably potential predictor of surgical outcome; then, a multivariate Cox proportional hazards regression model was applied with consideration of each significant risk factor. Statistical significance was evaluated using the log-rank test and comparison of the 95% confidence intervals (CIs) in Kaplan-Meier survival analysis. P-values < 0.05 indicated statistical significance, and P-values > 5% but <10% were reported to reflect trends. All statistical analyses were performed using SPSS 20.0 (IBM, USA).

Clinical Features
Demographics Our cohort consisted of 55 male and 28 female children. The mean age of seizure onset was 1.9 years (standard deviation [SD] 2.2, range 0.0-8.7), the mean duration of epileptic course prior to surgery was 3.1 years (SD 2.6, range 0.2-13.8), and the mean age of surgery was 5.0 years (SD 3.4, range 0.8-14.0).

Peri-Insular Hemispherotomy Surgery
Of 83 procedures, 46 were left-sided and 37 were right-sided. One patient experienced focal cortical resection prior to peri-insular hemispherotomy surgery due to intracerebral hemorrhage. In our cohort, no perioperative mortality was observed and APOS were noted in 7 patients (8.4%).

Surgical Outcomes
The mean follow-up period was 3.0 years (SD0.8, median 2.9 years). Twenty-one children (25.3%) were followed up for more than 4 years; 28 (33.7%) had a follow-up period between 3 and 4 years; and 34 (40.9%) were followed up for 2 to 3 years. At the last visit of follow-up (n = 83), 69 children (83.1%) were seizure-free after surgery (Engel class I) and 14 (16.9%) patients experienced seizure recurrence, including one with late remission, one with Engel class II, four with Engel class III, and five with Engel class IV outcomes. The seizure-free rate was 84.3% (70/83) upon the last follow-up.
Univariate Analysis ( Table 1) To explore the predictors, the patients were divided into 2 groups: seizure-free group and non-seizure-free group. Ignoring the impact of variables such as follow-up duration on the whole model, a non-lateralized interictal EEG pattern (P = 0.006), bilateral PET abnormalities (P = 0.001), and APOS (P = 0.01) were statistically correlated with seizure recurrence after surgery (P-values all <0.05) in univariate analysis.
Taking follow-up duration into consideration, a nonlateralized interictal EEG pattern (P = 0.038), bilateral PET abnormalities (P < 0.001), and APOS (P = 0.003) were also correlated with postoperative seizure recurrence using Log-Rank test in Kaplan Meier analysis.
By means of multivariate Cox proportional hazard analysis (Supplemental Table 2), bilateral PET abnormalities tended to independently predict poor prognosis of surgery (P = 0.001), and the patients with bilateral PET abnormalities were nearly 14 times more likely to undergo a failed peri-insular hemispherotomy (OR = 13.99, 95%CI = 2.75-71.17, P = 0.001).

DISCUSSION
In this study, we investigated the postoperative outcomes of 83 eligible patients with refractory epilepsy, hemisphere resection and postoperative follow-ups of at least 2 years. Among them, the seizure-free rates were 86.1% at 2 years, 82.0% at 3 years and 83.3% at 4 years and beyond. The mean duration of the followup was 3.0 years (2-4.5 years), with 83.1% (69/83) of the patients achieving seizure-free status. The seizure-free rates were similar to previously reported rates ranging from 63 to 80% in 5 large series (n = 84-170) (3-7). One longitudinal seizure outcome study showed that long-term outcomes change minimally over 2 to 11 years after surgery (3), which may be attributable to the selection criteria and the definition of a good prognosis for epilepsy surgery. Our study found that the seizure-free rate following hemispherical surgery decreased rapidly within 2 years (−3.1%) and decreased slightly over 2 years (−2.7%). However, the short follow-up duration was a limitation of this study and should be extended in further studies.
In this study, both of univariate and multivariate analyses methods were applied to explore the predictive factors of seizure outcomes. In the univariate analysis, a non-ipsilateralized interictal EEG pattern, bilateral PET abnormalities, and APOS appeared to potentially predict seizure recurrence after surgery (P < 0.05). In the multivariate analysis, bilateral PET abnormalities were the only predictor independently correlated with the unfavorable outcome of epilepsy surgery (P-values was 0.019 or 0.001).

Interictal EEG Pattern
One meta-analysis suggested that lateralized findings on interictal (OR = 1.66, 95%CI = 1.03-2.67, P = 0.04) or ictal (OR = 1.88, 95% CI = 1.15-3.07, P = 0.01) EEG may correlate with favorable seizure outcomes (11). The results of this study coincided with those of some previous studies. However, some other studies showed no correlation between interictal or ictal EEG and seizure outcomes (7). As far as seizure semiology was concerned (3,7), the presence of generalized seizures was related to unfavorable seizure outcomes (OR = 1.84, 95% CI = 1.18-2.89, P = 0.008). It is reported that the immature structural and functional brains of younger children are more likely to experience generalized seizures than adult brains (12,13). However, we did not determine any correlation between the presence of generalized seizures and unfavorable prognosis, although younger children constituted a large proportion of the cases in this study.

APOS
Acute seizures within 1 week after surgery were considered as an indication of postoperative epilepsy recurrence. In a large series of 170 children who underwent peri-insular hemispherotomy, 19% (31/163) of the children experienced APOS, and 71% (22/31) of them had poor postoperative seizure outcomes (7). Another study on seizure outcomes after extratemporal resections and hemispherectomy suggested consistent results that APOS could predict a poor long-term seizure outcome (14). In our study, 8.4% (7/83) of the patients had APOS, and a similar correlation between APOS and seizure recurrence was determined in univariate analysis (P = 0.027). But in the multivariate analysis, APOS could not independently predict seizure outcomes.

Bilateral PET Abnormalities
Only one study with a large sample size of hemispherectomy patients focused on bilateral PET abnormalities (7). The study revealed that bilateral PET abnormalities were associated with unfavorable outcomes (OR = 2.53, 95%CI = 1.02-5.85) and that they may be more reliable markers of potential independent epileptogenicity than bilateral structural abnormalities on brain MRI (7). In another study of 8 children with hemimegalencephaly, bilateral PET abnormalities were also proved to be a predictor for unfavorable seizure outcomes (15). In contrast, in a small sample of 18 infants who underwent hemispherectomy, no correlations were identified between poor seizure outcomes and bilateral It is suggested that patients with preoperative bilateral PET abnormalities should be carefully evaluated before hemispherotomy surgery and closely followed up after the operation. Besides the potential predictors mentioned above, other related factors such as contralateral MRI anomalies (8), etiology (6) and epileptic surgical procedure (10) were reported to be probably associated with poor seizure outcome.

The Contralateral Anomaly of Cerebral MRI
The value of preoperative contralateral MRI anomalies to predict the seizure outcomes after hemispherectomy therapy for epileptic patients were explored by a number of studies with inconsistent results. A study enrolling 110 patients who underwent hemispherectomy displayed no obvious correlation between contralateral MRI anomalies and postoperative seizure recurrence (8); while another research involving 43 patients reviewing hemispherectomy therapy suggested that contralateral MRI anomalies could predict the postoperative epilepsy recurrence (4). In our study, no correlation was identified between bilateral MRI abnormalities and postoperative epilepsy outcomes. Contralateral MRI abnormalities were foci of the preoperative evaluations before hemispherectomy. The structural abnormalities displayed in brain MRI, such as cortical malformations, abnormal white matter signals, abnormal white matter myelination and encephalomalacia, were supposed to be associated with poor seizure outcomes; but further investigations were still needed because of the limited number of relevant studies. In our opinion, contralateral MRI abnormalities were not an absolute contraindication for hemispherectomy therapy. Surgical candidacy should be determined by utilizing a combinative evaluation protocol including semiology analysis, preoperative vEEG, PET-CT as well as neuropsychological assessment. Emergent or early-stage MRI was helpful for determining the location and extent of lesions at early stage in favor of explaining clinical manifestations, especially when the patients had acquired etiology such as encephalitis and hemorrhage.

Etiology
One earlier research study found that patients with malformations of cortical development may have less favorable outcomes (6), but this has not been observed in other studies (3,5,7). In our study, different results were found in seizure outcomes of etiology. Progressive etiology had the best prognosis (100%, 16/16), followed by developmental etiology (90%, 18/20), and acquired etiology (80%, 32/40) was the worst. Possible explanations include different criteria for patient selection and various surgical techniques in different epilepsy centers. In a Cleveland epilepsy center, approximately 11.2% (19/170) of patients had undergone prior focal cortical resection (3). The integrity of excision or hemispherical disconnection also affects the surgical prognosis.

Epilepsy Surgery
In a previous study (10), the surgical technique was believed to be unrelated to the postoperative seizure outcomes (OR = 1.51, 95% CI 0.96-2.36, P = 0.07). As mentioned above, all children in our study underwent peri-insular hemispherotomy by the same chief surgeon, and only one patient had received prior focal cortical resection. The surgeries underwent by the same senior epileptic surgical specialist could eliminate the individual variations of surgical technique; and the works of a single epileptic center during a short period could reduce deviations produced by the different evaluative conception and the update of medical equipment. No perioperative or postoperative mortality was observed. However, we did not review or analyze surgical techniques, prior surgeries or other complications.

CONCLUSION
Hemispherotomy was an effective and safe surgical option to relieve the seizure burden of candidates with intractable epilepsy. The seizure-free rates after hemispherotomy were approximately 84.3% at a mean follow-up of 3 years. This study will facilitate better candidate selection for hemispherotomy and early identification of unfavorable seizure outcomes.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript/Supplementary Files.