Hemispherectomy in adults and adolescents: Seizure and functional outcomes in 47 patients

Summary

Objective:

To examine longitudinal seizure and functional outcomes after hemispherectomy in adults and adolescents.

Methods:

We reviewed 47 consecutive patients older than 16 years of age who underwent hemispherectomy between 1996 and 2016 at our center. Clinical, EEG, imaging, neuropsychological, surgical, and functional status data were analyzed.

Results:

Thirty-six patients were aged 18 or older at surgery; 11 were aged between 16 and 18 years. Brain injury that lead to hemispheric epilepsy occurred before 10 years of age in 41 (87%) patients. At a mean follow up of 5.3 postoperative years (median 2.9 years), 36 (77%) had Engel class I outcome. Longitudinal outcome analysis showed 84% seizure freedom (Engel IA) at 6 months, 76% at 2 years, and 76% at 5 years and beyond with stable longitudinal outcomes up to 12 years from surgery. Multivariate analysis demonstrated that acute post-operative seizures and contralateral interictal spikes at 6 month follow-up EEG were associated with seizure recurrence. Patients who could walk unaided pre-operatively and had no cerebral peduncle atrophy on brain MRI were more likely to experience worsening of motor function post-operatively. Otherwise, post-operative ambulatory status and hand function were unchanged. Of the 19 patients who completed neuropsychological testing, 17 demonstrated stable or improved post-operative outcomes.

Introduction

Hemispherectomy is an effective epilepsy surgery in children with medically refractory epilepsy secondary to large unilateral hemispheric epileptogenic lesions, with reported long term seizure freedom rates around 66–80%^1–6. Children selected for hemispherectomy typically have a pre-existing hemiparesis and their affected hemisphere is non-dominant for language, so hemispherectomy in this setting rarely leads to new unacceptable deficits. Children who walked prior to surgery retain or regain their ambulatory abilities after surgery⁷. Language and cognitive dysfunction may exist prior to surgery but postoperative regression in language skill is seldom seen. Because the ability of the opposite hemisphere to develop language function steeply declines when brain injury occurs after about 5 years of age, early surgery is often advocated in order to maximally exploit the window of neural plasticity, and enable optimal restoration of cognitive, motor, and particularly language function⁸. In a single center series of 170 children who underwent hemispherectomy, 95% had their surgery before 15 years of age¹.

There is limited information on outcome after hemispherectomy performed in adolescents and adults because this procedure is rarely performed in adults with refractory epilepsy. Concerns regarding risk to language and cognitive dysfunction, and motor weakness after hemispherectomy may be limiting the application of this surgical procedure in these patients. Most adult hemispherectomy series are small, with limited information on factors affecting functional and seizure outcome^9,10. Identification of prognostic factors that influence the seizures and functional outcome would aid in preoperative counseling. Additionally, longitudinal seizure outcome over several years has not been previously documented.

In this report, we describe our series of 47 patients who were 16 years or older at the time of hemispherectomy. This older cohort differs from pediatric hemispherectomy series in their etiology and electroclinical features, as a higher percentage of patients in our study have acquired lesions such as perinatal stroke. Thus, children with brain malformations and those with epileptic encephalopathy may undergo hemispherectomy at a younger age and are not as well represented in this series compared to pediatric studies. In this study, we report long-term seizure and functional outcomes and their predictors in all 47 patients as well as neuropsychological outcomes in a subset of 19 patients who completed comprehensive evaluations before and after surgery.

Methods

Patient Selection and Data Collection

This study was approved by our Institutional Review Board. From our epilepsy surgery database of 4,491 surgeries, we identified 356 patients who underwent hemispherectomy between 1996 and 2016. Of these, 47 were aged 16 years or older at the time of surgery. Twelve of the 47 patients were included in our prior study of 170 pediatric hemispherectomy patients.¹

We collected and analyzed preoperative demographic, imaging, and electroencephalographic data. Patients undergoing hemispherectomy typically had large hemispheric lesions, multifocal or poorly localized epilepsies, and/or pre-existing motor deficits with or without visual field defects. Language lateralization was performed in selected patients using functional MRI and/or Wada testing.

We have previously defined the terms functional hemispherectomy, modified anatomical hemispherectomy, and anatomical hemispherectomy used at our institution.¹ Information regarding previous surgeries such as focal resections, hemispheric surgery or vagal nerve stimulator placement was collected and recorded. For the purposes of this study, the most recent hemispherectomy was considered to be the surgery for which complications and outcomes would be recorded. For example, a patient who had a prior functional hemispherectomy but returned for an anatomical hemispherectomy would have their seizure outcome assessed in relation to the anatomical surgery. Operative time, complications, length of stay, and presence of acute post-operative seizures (defined as within 7 days of surgery) were collected and analyzed.

Functional status was assessed by categorizing hemiparesis, fine finger movements, ambulatory status, visual function, and language ability. All assessments of functional status were then categorized post-operatively as either unchanged, declined, or improved at the last follow-up visit with details on the specifics of each patient’s functional status given for context.

Seizure outcome data was obtained through our internal Epilepsy Center Outcomes Registry. Our primary measure of seizure outcome used the Engel classification. Patients were classified as Engel class I, II, III, or IV at last follow-up. Dates of seizure recurrence and last follow up were captured to allow a longitudinal analysis of outcomes. All patients with seizure recurrence had volumetric MRI sequences to check for incomplete disconnection. These were reviewed by the surgeon and an expert epilepsy neuroradiologist for residual interhemispheric or subcortical connections. Additionally, video EEG was performed to look for ictal patterns over the operated hemisphere with concurrent clinical signs of seizures. If documented, this is typically an electrophysiological marker of incomplete disconnection. Our approach to such cases has been outlined in a previous study.¹¹

Neuropsychological Outcomes

A subset of 19 patients (right =11, left =8) completed comprehensive neuropsychological evaluations prior to and approximately 10 months (mean 10.8, median 7.3, SD 7, range 6–33 months) following epilepsy surgery as part of standard clinical care. Because only a subset of patients completed testing, we examined demographic, operative and functional status differences between the patients who completed testing and those who did not. Patients who completed testing were generally higher functioning; they were significantly more likely to walk unaided, with less severe motor dysfunction and normal language function. Post-operatively, the patients who completed testing had a lower mean seizure frequency; there were no significant differences in Engel outcome between the two groups (Supplemental Table 1).

Most neuropsychological evaluations included measures of intellectual functioning, attention/working memory, language, memory, and executive functioning. Change scores were calculated for each cognitive measure and classified as “declined,” “no change,” or “improved” using established reliable change indices (90% confidence interval) for epilepsy that adjust for test-retest reliability and practice effects.^12–14 For details on the specific tests used as well as statistical analysis for these outcomes, see the Supplemental Information.

Statistical Analysis

All of the demographic data were summarized with descriptive statistics for each variable. Kaplan-Meier survival analysis was first used to calculate the probability of seizure freedom in the overall group before any outcome predictor analysis, and later by considering each of the significant risk factors. This allowed identification of potential prognostic indicators. When examining seizure outcome predictors, we categorized the patients into two groups, Engel I and Engel II-IV. For univariate analysis of categorical variables, χ² values were calculated, and Pearson test was used for hypothesis testing. If there were only 2 categorical variables used, Fisher’s exact test was used for hypothesis testing. For univariate analysis of continuous variables, one way ANOVA was used for hypothesis testing. A p-value of <0.05 was considered statistically significant for inclusion into the multivariate model. For the multivariate seizure outcome analysis, we constructed two Cox proportional hazard models: one consisting solely of preoperative predictors and one consisting of all potential predictors. Variables with a p-value of <0.05 in the multivariate model were then considered statistically significant. For language testing predictors, we first used univariate analyses of relevant variables including age at surgery, age of initial brain injury, operative side and presence of preoperative language abnormalities. For the multivariate analysis, a nominal logistic fit model was used including all of the relevant variables, and odds ratios were calculated with 95% confidence intervals, and p-values calculated based on probability> χ². All statistical analyses were performed using JMP 13.0 (SAS Institute, Cary, NC).

Results

Demographics, seizure burden and imaging characteristics

Forty-seven patients aged 16 or older underwent hemispherectomy between 1996 and 2016. Preoperative clinical, imaging and EEG variables are shown in Table 1. Of 47 patients, 20 were 25 years or older at the time of surgery; 16 were aged between 18 and 24 years, and 11 were aged between 16 and 18 years. The mean age of initial evaluation at our institution was 24.7 years and the mean age of surgery was 26.4 years. The estimated age at acquisition of the epileptogenic brain lesion was as follows (i) prenatal, perinatal, or before 2 years of age in 36 patients (76%); (ii) between 3 and 10 years of age in 5 (11%); (iii) between 11 and 16 years of age in 4 (9%), and (iv) after 16 years of age in 2 patients (4%). The most common etiology was perinatal infarction (47%), bringing the median age of lesion acquisition to 1 month of age. Malformations of cortical development constituted 15%, while 30% had later-acquired conditions, such as Rasmussen encephalitis, trauma, and encephalitis. Seizure burden was high, with 93% having either daily or weekly seizures.

Table 1.

Univariate analysis of preoperative clinical, imaging, and EEG findings in relation to seizure outcome

	Overall N = 47	Engel I N = 36	Engel II-IV N = 11	p value
Gender, M:F	23:24	17:19	6:5	0.67
Side of surgery, L:R	17:30	13:23	4:7	0.99
Median age at surgery, years (range)	22 (16.2–61)	21.2 (16.2–61)	23.7 (16.6–56)	0.22
Median age at initial brain insult, years (range)	0.08 (0.01–12)	0.1 (0.1–17.8)	0.75 (0.1–40)	0.13
Median age at seizure onset, years (range)	3 (0.1–42)	1.9 (0.1–21)	3 (1–42)	0.06
Median duration of epilepsy, years (range)	17.3 (2.1–52.6)	17.4 (2.1–51)	15.1 (8–52.6)	0.58
Number of seizure types, n (%)				0.24
Single	20	17 (85%)	3(15%)
Multiple	27	19 (70%)	8 (30%)
Seizure types, n (%)				0.34
Exclusively focal	26	18 (69%)	8 (31%)
Exclusively generalized	1	1 (100%)	0 (0%)
Focal and generalized	20	17 (85%)	3 (15%)
Seizure frequency, n (%)				0.22
Daily	25	16 (64%)	8 (32%)
Weekly	18	16 (89%)	2 (11%)
Monthly	4	3 (75%)	1 (25%)
Mean # of failed AEDs, n (±SD)	8.2 (± 3.1)	7.7 (±2.6)	10.1 (±4.2)	0.04
Mean # of preoperative AEDs, n (±SD)	3.0 (± 1.0)	3.2 (± 0.9)	2.3 (± 0.7)	0.005
Etiology, n (%)				0.14
Perinatal stroke	22	18 (82%)	4 (18%)
Malformation of cortical development	7	5 (71%)	2 (29%)
Rasmussen’s encephalitis	6	5 (83%)	1 (17%)
Traumatic brain injury	5	5 (100%)	0 (0%)
Other Encephalitis	4	2 (50%)	2 (50%)
Other	3	1 (33%)	2 (67%)
Interictal EEG, n (%)				0.75
psilateral spikes only	24	19 (79%)	5 (21%)
Bilateral or generalized Spikes	17	12 (71%)	5 (29%)
No spikes	6	5 (83%)	1 (17%)
Lateralization on ictal EEG, n (%)				0.21
Ipsilateral to lesion	34	25 (74%)	9 (26%)
Non-lateralized	6	5 (83%)	1 (17%)
No EEG change or no seizures	5	5 (100%)	0 (0%)
Contralateral to lesion	1	0 (0%)	1 (100%)
Unavailable	1	1 (100%)	0 (0%)
Contralateral MRI findings, n (%)				0.90
Normal	31	25 (81%)	6 (19%)
Subcortical white matter abnormalities	6	4 (67%)	2 (33%)
Cortical abnormality limited to 1 lobe	4	3 (75%)	1 (25%)
Cortical abnormality in ≥ 2 lobes	4	3 (75%)	1 (25%)
Unavailable	2	1 (50%)	1 (50%)
PET findings (n=25), n (%)				0.67
Ipsilateral abnormalities only	20	16 (80%)	4 (20%)
Bilateral abnormalities	5	3 (60%)	2 (40%)

Pre-operative language lateralization

Thirty-three patients (70%) did not undergo any type of pre-operative testing for language localization (Supplemental Table 2). Of the fourteen patients who underwent language testing, three underwent only fMRI that indicated language localization in the contralateral hemisphere. Seven patients underwent only Wada testing, and all of these patients had surgery prior to 2005 when fMRI was not typically used in determining language function. Six of these patients had testing indicating contralateral language, and one patient had bilateral language representation. Four patients had both Wada testing and an fMRI, typically because the fMRI was difficult to interpret or appeared to show bilateral language representation. Three of these patients had confirmed contralateral language on Wada testing while one was unable to complete the testing due to a complication (hemiparesis resulting from a small internal capsule infarct reported¹⁵).

Univariate analysis of the factors associated with language testing revealed that older age at surgery and patients undergoing left-sided hemispherectomies were significantly more likely to have some type of language testing. Using multivariate analysis, only left-sided hemispherectomy was associated with preoperative language testing (Supplemental Table 2).

Operative details, complications and pathology

Details on the types of surgery performed, pathology, and post-operative complications are shown in Table 2. In patients who had a prior focal resection, the latency between first and last surgeries was a mean of 10.7 years. The reason for initial focal resection could not be determined reliably by chart review. Of the 10 anatomic or modified anatomic hemispherectomies, 8 had persistent seizures despite prior functional hemispherectomy either at our institution or elsewhere.

Table 2.

Univariate analysis of operative variables, pathology and complications as related to seizure outcome

	Overall N = 47	Engel I outcome	Engel II-IV outcome	P value
Operative side: Right, n (%)	30	23 (77%)	7 (23%)	0.99
Type of hemispherectomy, n (%)				0.76
Functional	37	28 (76%)	9 (24%)
Modified anatomic	1	1 (100%)	0 (0%)
Anatomic	9	7 (78%)	2 (22%)
Prior surgery, n (%)a
Hemispheric	8	6 (75%)	2 (25%)	0.90
Focal	14	7 (50%)	7 (50%)	0.0067
Vagal nerve stimulator	9	3 (33%)	6 (67%)	0.0015
No prior surgery	25	23 (92%)	2 (8%)	0.008
Mean operative time, min (± SD)	241 (±93)	231 (±91)	291 (±109)	0.45
Mean length of stay, d (± SD)	9.9 (±8)	7.8 (±3.2)	17.1 (±14.4)	0.06
Pathology, n (%)				0.10
Gliosis	25	17 (68%)	8 (32%)
Gliosis with features of cortical dysplasia	8	6 (75%)	2 (25%)
Malformation of cortical development	7	7 (100%)	0 (0%)
Rasmussen’s encephalitis	3	3 (100%)	0 (0%)
Other	4	3 (75%)	1 (25%)
Complications, n (%)				0.15
None	30	25 (83%)	5 (17%)
Aseptic meningitis	10	9 (90%)	1 (10%)
Hydrocephalus	5	3 (60%)	2 (40%)
Infectionsb	5	2 (40%)	3 (60%)
Meningitis	2	0 (0%)	2 (100%)
Wound	2	0 (0%)	2 (100%)
Urinary tract infection	3	2 (67%)	1 (33%)
Deep vein thrombosis	4	4 (100%)	0 (0%)
Intracerebral hemorrhage	1	0 (0%)	1 (100%)
Acute post-operative seizures, n (%)	6	2 (33%)	4 (67%)	0.01
6-month post op EEG: spikes in contralateral hemisphere	2	0 (0%)	2 (100%)	0.005

^a.Some patients had multiple types of surgery and therefore these numbers add up to more than 47.

^b.Two patients had wound infections and meningitis so these numbers add up to more than 5.

Seizure outcomes

At a mean follow-up of 5.4 years (median 2.9 years), 36 of 47 patients (77%) had an Engel I outcome. These outcomes remained stable up to twelve years after surgery. Of these, 27 have been completely seizure-free since surgery (Engel IA) while nine were class IB-D. Five patients (10%) were in Engel class II and III while 6 patients (13%) were in Engel class IV. The Kaplan-Meier curve in Figure 1 demonstrates longitudinal seizure recurrence rates. Of the 27 Engel IA patients, four were completely off AEDs at last follow-up. On average, the Engel IA patients were taking 3.2 AEDs preoperatively, and 1.2 AEDs postoperatively. Of the 9 Engel Class 1B-D patients, five achieved remission for at least 1 year at last follow up. Seizure recurrence appeared related to anti-epileptic drug (AED) withdrawal in three patients. One patient had one seizure after being seizure-free for two years and was weaned off of AEDs, and another patient had seizure recurrence 8 years after surgery. Both patients regained seizure freedom on resuming the AED. The third patient missed doses of AEDs around 4–5 months postoperatively and the long term seizure freedom status is unknown.

Figure 1:

Longitudinal seizure outcome after hemispherectomy in 47 adults and adolescents. Kaplan-Meier survival curves illustrate the seizure freedom after hemispherectomy after various time intervals.

Functional status outcomes

Pre- and post-operative functional outcome in motor, language and visual function are shown in Table 3.

Table 3.

Pre- and post-operative functional status in 47 adult hemispherectomy patients

PRE-OPERATIVE N = 47	N (%)	POST-OPERATIVE N = 47	N (%)
HEMIPARESIS		HEMIPARESIS
None	1 (2%)	Unchanged	28 (60%)
Mild to moderate	36 (77%)	Worse	16 (34%)
Severe	6 (13%)	Better	3 (6%)
Asymmetric quadriparesis	4 (9%)
FINE FINGER MOVEMENTS		FINE FINGER MOVEMENTS
None	29 (62%)	Unchanged	35 (75%)
Minimal function, helper hand	11 (23%)	Worse	11 (23%)
Fair dexterity	4 (9%)	Unknown	1 (2%)
Normal	2 (4%)
Unknown	1 (2%)
AMBULATORY STATUS		AMBULATORY STATUS
Walks unaided	33 (70%)	Unchanged	33 (70%)
Uses aides/orthoses to walk	10 (21%)	Walks using new aides/orthoses	13 (28%)
Few steps with assistance	0 (0%)	Better than pre-op	1 (2%)
Non-ambulatory	4 (9%)
LANGUAGE		LANGUAGE
Normal	24 (51%)	Grossly Unchangeda	46 (98%)
Delayed	18 (38%)	Worseb	1 (2%)
Non-verbal	4 (9%)
Acquired aphasia	1 (2%)
VISUAL FUNCTION		VISUAL FUNCTION
Present	25 (53%)	Unchanged	20 (43%)
Normal	10 (21%)	Worse	21 (45%)
Unavailable	10 (21%)	Unavailable	6 (13%)
Suspected subjectively	2 (4%)

^a.This categorization was based on the neurological exam findings in post-operative visits. Details on language and cognitive function can be found in the neuropsychological outcomes in Figure 2.

^b.This case has been reported in detail in 2004. Loddenkemper T, Dinner DS, Kubu C, et al. Aphasia after hemispherectomy in an adult with early onset epilepsy and hemiplegia. J Neurol Neurosurg psychiatry 2004; 75:149–151.

Motor:

Post-operative motor status was categorized as unchanged, worse or better, as shown in table 3. In 60% of patients, hemiparesis was unchanged after surgery; a third of patients who worsened after surgery typically had mild weakness pre-operatively. Rare patients who were considered “improved” typically had a severe baseline deficit with reduction in their spasticity post-operatively. All patients in whom follow up was available were able to walk independently, some requiring a walking aides or orthoses. One patient could only take a few steps with assistance at last follow-up 15 months post-surgical admission. One patient who was non-ambulatory preoperatively improved significantly postoperatively such that she could ambulate with assistance postoperatively.

Language:

Regardless of the level of preoperative language function, all patients, except one, had unchanged language function post-operatively. This patient, reported elsewhere¹⁶, had an expressive aphasia postoperatively that did not improve even though pre-operative Wada testing demonstrated preserved language function after left-sided injection.

Neuropsychological testing:

Patients who completed pre- and postoperative neuropsychological testing were 29 years old on average (±12 years) with 12 years of education (±2 years). Of the 19 patients who completed neuropsychological testing before and after surgery, 11 patients (58%) did not show cognitive decline on any cognitive measure and 6 patients (32%) declined in only 1 cognitive domain. A significantly larger proportion of patients (31%) demonstrated postoperative improvement on a novel problem-solving measure than expected by chance alone (Figure 2). Specifically, a substantial proportion of patients made fewer perseverative errors on this task following surgery (Wisconsin Card Sorting Test – Perseverative Errors) as compared to their preoperative performance. The proportion of patients demonstrating significant cognitive postoperative change (in either direction) did not exceed chance for any of the other cognitive measures examined. Two patients (11%) demonstrated reduced scores in 3 or more cognitive domains. One of these patients was the patient who experienced post-operative aphasia despite pre-operative Wada testing indicating supported language function in the opposite hemisphere.

Figure 2:

Cognitive change after hemispherectomy. Clinically meaningful cognitive change was determined using published Reliable Change Indices for epilepsy with a 90% confidence interval.^11–13 Values in parentheses represent the sample size for each measure. Significant findings suggest a larger than expected proportion of patients demonstrated clinically significant change (improvement or decline) following surgery than expected by chance.

Seizure and functional status outcome predictors

There were a number of factors associated with poor seizure outcome at last follow-up on univariate analysis including more failed anti-seizure drugs preoperatively, prior focal resective surgery, prior vagal nerve stimulator (VNS) placement, acute post-operative seizures, and contralateral interictal spikes on EEG at six month follow-up, as shown in Tables 1 and 2. A Cox proportional hazard model including only preoperative variables demonstrated that only the presence of a prior VNS correlated with poor outcome. A model including both pre- and post-operative variables showed that contralateral interictal EEG spikes at 6 month follow-up and acute post-operative seizures were predictive of poor outcome at last follow-up (Table 4).

Table 4.

Predictors of post-operative seizure outcome in two multivariate models

Model 1a: Only pre-operative predictors of seizure freedom	Multivariate analysis Risk Ratio (95% CI), p value
Prior VNS placement	0.04 (0.01–0.53), p=0.01
Number of pre-op AEDsb	2.48 (0.83–8.85), p=0.11
Prior focal surgery	0.21 (0.02–1.93), p=0.17
Number of total failed AEDsb	1.09 (0.84–1.47), p=0.55
Model 2c: All potential predictors of seizure freedom
Contralateral interictal spikes on EEG at 6 months post-op	0.01 (0.001–0.21), p=0.0066
Acute post-operative seizures	0.06 (0.01–0.93), p=0.04
Prior VNS placement	0.12 (0.01–6.8), p=0.27
Number of pre-op AEDsb	1.35 (0.34–4.55), p=0.62
Prior focal surgery	0.71 (0.07–10.2), p=0.77

^a.Whole model p test = 0.0038

^b.Risk ratio is expressed as risk per unit change in regressor (1 medication)

^c.Whole model p test = 0.001

When we examined predictors of motor function outcomes, we generally found that better preoperative motor function predicted worsened post-operative outcome. This typically correlated with the severity of cerebral peduncle atrophy on pre-operative MRI such that patients with less cerebral peduncle atrophy were more likely to have worsened post-operative function (Supplemental Table 3). We found no predictors of neuropsychological outcomes.

Discussion

In this series of hemispherectomy in 47 adolescents and adults, 77% of patients were Engel class I at their most recent follow-up (median 2.9 years, mean 5.3 years). Longitudinal outcome analysis showed 84% seizure freedom (Engel IA) at 6 months, 77% at 2 years and 77% at 5 years and beyond with stable longitudinal outcomes up to 12 years from surgery. Seizure outcomes are comparable to Engel class I outcome of 55–88% in prior studies.^17–20 Series with higher rates of acquired pathologies, as in this cohort, tend to have higher seizure freedom rates.^6,17,19–20 These outcomes are also similar to pediatric hemispherectomies with 73% overall seizure freedom in a pooled analysis of 1,161 pediatric patients⁶. Longitudinal outcome of pediatric hemispherectomy series from our center showed comparable seizure freedom rates of 78% at 6 months, 71% at two years and 63% at five years and beyond.¹

Predictors of seizure outcome:

When we analyzed both pre-and post-operative variables, acute post-operative seizures, and interictal spikes in the contralateral hemisphere at the six month follow-up were predictive of seizure outcome.

Seizures in the acute post-operative period are often viewed as transient due to peri-operative factors including edema, blood products in the brain, and/or electrolyte abnormalities. However, as shown in this and other studies, it is frequently a sign of early failure of surgery, particularly in patients with high pre-operative seizure burden.^1,21,22 EEG at 6-months post-surgery was also a powerful tool to predict the likelihood of seizure freedom. None of the patients with Engel class I outcome had spikes in the contralateral hemisphere post-operatively. In patients with persistent seizures after hemispherectomy, the presence of epileptiform discharges in the opposite hemisphere may potentially indicate independent epileptogenicity from the un-operated hemisphere. On the contrary, bilateral or generalized epileptiform discharges on pre-operative EEG was noted in a third of the patients who were seizure-free after surgery and thus these features should not be used alone to exclude patients from surgery. Epileptiform discharges on the operated hemisphere are a common finding on post-op EEG, and have no clinical relevance in a patient who is clinically seizure free, as residual disconnected tissue may continue to express epileptiform discharges. Small sample sizes of prior adult hemispherectomy series⁹ have not allowed for identification of such seizure outcome predictors. Prior pediatric studies have variably identified underlying etiology^2,4,6, ‘unambiguous’ MRI abnormalities in contralateral hemisphere,²³ bilateral FDG-PET abnormalities,¹ acute postoperative seizures,^1,2 and contralateral spikes on follow up EEG¹ as potential predictors of seizure recurrence. Etiology,¹ and presence of MRI abnormalities^1,24 in the contralateral hemisphere did not affect outcome in some studies.

When we only analyzed pre-operative variables in a multivariate model, patients with VNS had poor seizure outcome. As a group, patients with VNS tended to have more severe hemiparesis or asymmetric quadriparesis and non-ambulatory, failed more medications, and had contralateral interictal EEG spikes. Thus, in this series, VNS implantation was likely a surrogate marker for more severe disease, both in terms of patients’ functional status, and epilepsy with a higher potential for bilateral epileptogenicity. However, 3 of the 9 patients who had prior VNS achieved Engel I outcome after hemispherectomy.

Motor function after hemispherectomy:

All patients in our series who could walk unaided preoperatively were able to maintain their ability to walk though some patients required some type of orthoses or aide post-operatively who did not previously use one. While most prior adult studies have not examined ambulatory status specifically, some studies have demonstrated small improvements in gait and there have been no reports of loss of ambulation.⁹ Studies in children have demonstrated similar outcomes with rare instances of worsening ambulation post-operatively.^5,7 Seizure recurrence and contralateral MRI abnormalities were correlated with poor motor outcome in children,⁷ but were not related to post-operative outcome in our study. Preoperative ambulatory status and the severity of cerebral peduncle atrophy on MRI were predictive of post-operative motor function. Patients who could walk unaided pre-operatively with no peduncle atrophy were much more likely to experience worsening post-operatively (Supplemental Table 3). Similar to ambulatory status, better preoperative hand function and mild or no cerebral peduncle atrophy on MRI predicted worsened hand function post-operatively (Supplemental Table 3). Preserved peduncle size likely represents partially preserved primary sensorimotor cortical projections that account for a significant portion of the peduncle. Conversely, patients with significant preoperative motor weakness have peduncle atrophy. In a prior study, children with larger cerebral peduncles on the side of surgery are more likely to have a worsened hemiparesis post-operatively²⁵. Thus, while performing hemispherectomy earlier in life is likely to lead to better motor outcomes, our study shows that it can still be safe to do so in adulthood depending on the MRI and pre-operative functional status of the individual. The impact of the functional deficits on the quality of life was not evaluated in our study. In one prior study, 17 of 20 patients reported improved quality of life, 2 reported ‘no difference’, and one experienced “deterioration.”¹⁹

Neuropsychological outcome:

In the 19 patients who had pre- and post-operative neuropsychological data, 17 patients did not show a significant decline in neuropsychological outcomes post-operatively. While some studies have demonstrated that adolescents with hemispheric epilepsy due to perinatal stroke with later epilepsy onset who have later surgery tend to have better seizure and cognitive outcomes^26,27, we did not have enough patients in this category to validate this finding. We did demonstrate a significant improvement in a novel problem solving measure in 31% of patients, owing to fewer perseverative errors. There were two patients who declined in multiple cognitive domains. One of these was the patient who had passed pre-operative Wada testing but had a significant aphasia post-operatively, reported earlier.¹⁶ The other appeared to be having significant cognitive side effects from one of his medications post-operatively. Thus, most patients who are able to complete neuropsychological testing have a reasonable expectation of being unchanged cognitively when undergoing hemispherectomy. Although reporting of neuropsychological outcomes is rare in this patient population, one prior study did demonstrate an 8 point improvement in full scale IQ on the Weschler Adult Intelligence Scale (WAIS-CR) post-hemispherectomy.²⁰ This study also showed that patients with right-sided hemispherectomies had a larger improvement in verbal IQ compared to left-sided hemispherectomies. In our study, there were no differences in testing between patients who had either left or right-sided hemispherectomies.

Thus, in our series of adult patients, the operated hemisphere was the non-dominant hemisphere for language, irrespective of the anatomical side. In addition, because the cohort of patients who completed testing were generally higher functioning (Supplemental Table 1) and therefore at higher risk of cognitive decline, it is particularly encouraging that this group did not worsen. We do note, however, that most patients (15/19) had their follow-up testing 6–12 months post-operatively (mean = 10.7 months, SD = +7 months) and thus we cannot comment on their long term neuropsychological outcomes. We were unable to determine the impact of post-operative complications and its impact on cognitive outcome. Of 7 patients with decline in at least one domain, 5 had no significant postoperative complications; one patient had hydrocephalus, and other had aseptic meningitis.

Language lateralization testing:

The majority (33 of 47) of patients in this series underwent hemispherectomy without the need for a language lateralization procedure, including seven left-sided hemispherectomies. None of the patients who did not undergo language lateralization testing had any unexpected post-operative worsening of language function. When we investigated the reasons that prompted a language evaluation, we found only one statistically significant predictor: operative side. Of 17 left-sided surgeries, 10 had language testing, whereas only 4 of 30 right-sided surgeries had language testing. In one patient who developed unexpected decline in language function, left intracarotid amobarbital test provided misleading results. The pre-operative “red flags” for possible language function on the left in this patient were (i) lesion acquisition at 5 years of age, which is at the limits of “safety” for assuming language transfer to the right hemisphere, and (ii) the predominantly subcortical location of his posttraumatic encephalomalacia. Functional MRI for language was not in practice at the time of surgery for this patient. Use of functional MRI and or bilateral Wada test may have provided evidence for language in both hemispheres in this patient.

Our study did not allow us to investigate the precise reasons for testing in selected patients. In our experience, if the brain injury occurred before 6 years of age⁸, and the extent of the hemispheric lesion is extensive involving the perisylvian regions, then that hemisphere is unlikely to harbor language function. Thus, most patients with right hemispheric disease with such features did not undergo language testing. If the age at injury was after age 6 years, and the extent of injury is patchy and spares some perisylvian regions, then language testing is strongly considered, especially with left hemispheric disease. Baseline language ability, feasibility of the language lateralization procedures such as fMRI/Wada, and the experience of the treating physicians may also have influenced the decision to test or not. It is difficult to make firm recommendations from our data as patients who were considered for hemispherectomy and later excluded from surgery based on language studies would not be captured in our study. Thus, when considering how best to reduce the incidence of post-operative language deficits in adult hemispherectomy candidates, patient selection supplemented with preoperative language lateralization when there is any uncertainty as to language laterality are probably the most important factors.

Comparison to pediatric hemispherectomy series:

The etiology of the epilepsy, seizure outcome, and functional outcome in this series are largely similar to those in the pediatric hemispherectomy experience. Epileptogenic lesions that cause hemispheric epilepsy in children such as multi-lobar malformations, large ischemic strokes, Rasmussen encephalitis, and encephalomalacia due to other causes (trauma, hemorrhage) were represented in adult series as well. Malformations were less frequent, accounting for 15% of causes in this series compared to 36–47% in pediatric series.^1–6 Large ischemic strokes and encephalomalacia due to other causes accounted for 63% of all etiologies in our series, compared to 22–46% in pediatric series.^1–6 Although we were unable to confirm etiology as a predictor of outcome in our study, patients with acquired etiologies likely have superior seizure freedom rates⁶ and so this could partially explain why our seizure outcomes are in the higher range of reported values. The risk for refractory epilepsy in patients with ischemic stroke appears to be an age dependent phenomenon, as large ischemic strokes acquired in early infancy are a relatively common substrate in hemispherectomy series in children and adults. In this series of 47 patients, 76% of patients sustained their primary brain injury before the age of 2 years and another 11% between age 2 and 10 years. This is similar to other series in which the majority of patients also sustained their injury in childhood; in one series of 27 patients,¹⁹ 24 had lesion onset before 16 years of age, and in another series of 25 patients all had disease onset before 15 years.²⁰ Thus, the majority of patients in these “adult hemispherectomy series” could be seen as pediatric hemispherectomy candidates who had their surgery in adulthood.^17–20 The similarities in seizure outcome and functional outcome between pediatric and adult series is therefore not surprising. The similarities in functional outcome also suggest that the neural plasticity and language transfer across hemispheres are determined by the age at brain injury rather than the age at surgery. A significant delay in referral for epilepsy surgery in this cohort was evident from the mean age of 25 years at the initial evaluation at our center. The expected negative impact on seizure outcome by the delay in this cohort may have been neutralized by the higher proportion of patients with early onset stroke who tend to have better outcomes⁶.

This study had several strengths and limitations. Because of the relatively larger number of patients, we were able to construct long term Kaplan-Meier curves in a meaningful fashion and search for potential prognostic factors for both seizure and functional outcomes. All 47 consecutive, eligible patients were included and no patient was lost to follow-up. We carefully documented the presence of abnormalities in the opposite hemisphere and the practice of language lateralization procedures. On the other hand, the study is retrospective in nature and therefore subject to a number of biases. The selection bias inherent in studies such as this one is partially evident in our analysis of language testing as highlighted earlier. Thus, the data on the need for language testing should be interpreted with caution. Second, only 19 of the 47 patients underwent neuropsychological testing, and so our neuropsychological findings may be limited to the patients whose functional status allows them to complete the testing. Third, due to the retrospective chart review study design, our ratings of functional status were limited to qualitative ordinal categories rather than objective semi-quantitative scales. The impact of the functional deficits on the quality of life was also not evaluated. Finally, our decision to include patients between 16 and 18 years of age in our study may raise some concerns. However, for assessing functional outcome after a procedure such as hemispherectomy, a conservative estimated age cutoff beyond which meaningful language plasticity is unlikely to occur was considered more meaningful than using an arbitrary age of 18 years.

In summary, this experience demonstrates that seizure outcomes in adults undergoing hemispherectomy are similar to the results seen in children. Acute post-operative seizures, and contralateral interictal spikes on 6 month follow up EEG are poor prognostic indicators and likely indicative of residual or contralateral epileptogenicity, respectively. While patients with preserved hand function and cerebral peduncle size on MRI will almost certainly experience post-operative worsening, ambulatory status and gross motor function are typically unchanged. Finally, adult hemispherectomy patients can expect their cognition to be relatively unchanged after surgery.

Key Point Box.

• Seizure outcomes in adults undergoing hemispherectomy are similar to the results seen in children

• Acute post-operative seizures and contralateral interictal epileptiform discharges on 6 month follow-up EEG are poor prognostic indicators

• Ambulatory status, gross motor function and cognition are typically unchanged post-operatively

• Patients with preserved hand function and cerebral peduncle size on MRI will likely experience post-operative worsening

Significance:

Hemispherectomy in adults is a safe and effective procedure, with seizure freedom rates and functional outcome similar to that observed in children.