Surgical Outcomes in Post-Traumatic Epilepsy: A Single Institutional Experience

Abstract

BACKGROUND

Post-traumatic epilepsy (PTE) is a debilitating sequela of traumatic brain injury (TBI), occurring in up to 20% of severe cases. This entity is generally thought to be more difficult to treat with surgical intervention.

OBJECTIVE

To detail our experience with the surgical treatment of PTE.

METHODS

Patients with a history of head injury undergoing surgical treatment for epilepsy were retrospectively enrolled. Engel classification at the last follow-up was used to assess outcome of patients that underwent surgical resection of an epileptic focus. Reduction in seizure frequency was assessed for patients who underwent vagal nerve stimulator (VNS) or responsive neurostimulator (RNS) implantation.

RESULTS

A total of 23 patients met inclusion criteria. Nineteen (82.6%) had mesial temporal sclerosis, 3 had lesional neocortical epilepsy (13.0%), and 1 had nonlesional neocortical epilepsy (4.3%). Fourteen patients (60.9%) underwent temporal lobectomy (TL), 2 underwent resection of a cortical focus (8.7%), and 7 underwent VNS implantation (30.4%). Three patients underwent RNS implantation after VNS failed to reduce seizure frequency more than 50%. In the patients treated with resection, 11 (68.8%) were Engel I, 3 (18.8%) were Engel II, and 2 (12.5%) were Engel III at follow-up. Average seizure frequency reduction in the VNS group was 30.6% ± 25.6%. RNS patients had reduction of seizure severity but seizure frequency was only reduced 9.6% ± 13.6%.

CONCLUSION

Surgical outcomes of PTE patients treated with TL were similar to reported surgical outcomes of patients with nontraumatic epilepsy treated with TL. Patients who were not candidates for resection demonstrated variable response rates to VNS or RNS implantation.

ABBREVIATIONS

CI

confidence interval

EEG

electroencephalogram

LNC

lesional neocortical

MRI

magnetic resonance imaging

NNC

nonlesional neocortical

PTE

post-traumatic epilepsy

RR

relative risk

RNS

responsive neurostimulator

TL

temporal lobectomy

TBI

traumatic brain injury

VNS

vagal nerve stimulator

Traumatic brain injury (TBI) affects approximately 2.8 million persons per year and constitutes a major cause of morbidity and mortality in the United States.¹ For those patients who survive the initial injury, post-traumatic epilepsy (PTE) is a significant contributor to the morbidity and poor quality-of-life associated with head injury. Reported rates of PTE can be up to 20% of patients with severe TBI.^2,3 The standard classification of post-traumatic seizures is organized by the timing of seizures with respect to the brain injury (immediate <24 h, early <1 wk, and late >1 wk), with each subset having its own set of outcomes.^2,4 Importantly, late seizures, the category typically equated with PTE, can develop even decades after initial injury, occurs in up to 42% of patients with severe head injury, and is not prevented by seizure prophylaxis during the acute period of head injury.^2,4-7 While medical management is the mainstay of PTE, this clinical entity can be particularly difficult to manage with antiepileptic drugs.^8,9 Part of the challenge of managing PTE is the underlying heterogeneity of the injury mechanisms leading to diverse neural injury patterns. Moreover, there is evidence in the literature supporting the existence of multiple distinct histopathologic subtypes of PTE that may influence outcome.¹⁰

Like other forms of epilepsy, surgical intervention may be indicated for selected patients who have medically refractory PTE. TBI is often considered a diffuse injury process, and this may lead to the assumption that PTE is not amenable to surgical treatment. The data supporting PTE as a multifocal process are sparse. Furthermore, neuromodulatory surgical therapies can be used to treat multifocal epilepsy. There exists class I evidence for anterior temporal lobectomy (TL) in the setting of mesial temporal lobe epilepsy, with reported seizure free rates of 58% compared to 8% for medically managed patients,¹¹ and these findings have been supported by follow-up pooled analyses.^12,13 Surgical outcomes in patients with PTE are somewhat conflicting within the literature, and surgical treatment of PTE is generally underreported with only several small case series published to date examining resective approaches in this population of patients.^10,14-19 We hypothesized that surgical outcomes in PTE may be similar to surgical outcomes in nontraumatic epilepsy. The present retrospective study describes the surgical outcomes for PTE patients following resection or implantation of a neuromodulatory device from a single institutional cohort.

METHODS

Patients undergoing surgical treatment for epilepsy at our institution from February 2003 to September 2015 were identified from a prospective registry of patients. Patients were excluded if they lacked at least 12-mo follow-up or had incomplete medical records. Our primary cohort included patients in whom the principal risk factor for epilepsy was traumatic head injury. Two patients included in the study had a history of febrile seizures, although the TBI was the most proximal risk factor for the onset of epilepsy. Within the registry of patients undergoing epilepsy surgery, patients with nontraumatic epilepsy who underwent temporal lobectomies were also identified and analyzed to serve as a control cohort for comparison of outcomes. Patient baseline demographics, trauma history, epilepsy-related clinical and treatment variables, postoperative Engel classification, and postoperative seizure frequency reduction were obtained through retrospective chart review. Seizure frequency reduction was calculated using preoperative and postoperative patient-reported seizure frequency. Patient-reported seizure frequency was recorded in the medical record as seizures per day, week, or month. For those that reported seizures per day or seizures per week, multiplication was used to extrapolate seizures per month. The study was approved by our Institutional Review Board. Patient consent was not sought because of the retrospective nature of this study.

Severity of head injury (mild, moderate, or severe) was defined as in Annegers et al.²⁰ Patients with mild injury were those without skull fracture and with less than 30 min of loss of consciousness or post-traumatic amnesia. Patients with moderate injury were those with skull fracture or loss of consciousness or post-traumatic amnesia of 30 min to 24 h. Patients with severe injury were those who had loss of consciousness or post-traumatic amnesia 24 h or greater, or patients who had a brain contusion, or intracranial hematoma.

All patients underwent standard epilepsy workup including magnetic resonance imaging (MRI) and scalp video-electroencephalogram (EEG) monitoring. Intracranial EEG monitoring was performed in select patients (n = 8) for which noninvasive localization was inconclusive. Strip, grid, and depth electrodes were used for intracranial monitoring in these cases.

Patient assessment and management was decided at an institutional, multidisciplinary epilepsy surgical case conference. For the patients included in the study, epilepsy was classified as either mesial temporal sclerosis (MTS), lesional neocortical (LNC) epilepsy, or nonlesional neocortical (NNC) epilepsy based on MRI findings. During the initial study period, available surgical treatment options consisted of standard TL and amygdalohippocampectomy, focal cortical resection, and vagal nerve stimulation (VNS). With the advent of responsive neurostimulation (RNS), patients in the study cohort who were not candidates for resection and who failed VNS therapy (seizure frequency reduction of less than 50%) were offered RNS system placement. Surgical procedures were performed by 1 of 2 surgeons (Gordon Baltuch or Timothy Lucas). Histology was performed by clinical neuropathologists within our health system. Means are reported with their respective standard deviations. Statistical comparisons between groups were performed with the Chi-square analysis, unpaired/two-tailed t-tests, and Fisher's exact test where appropriate. A P-value of <.05 was used as the threshold for statistical significance. All statistical analyses were performed using Prism (GraphPad, La Jolla, CA).

RESULTS

During the study period, a total of 389 patients underwent surgery for treatment of epilepsy. Of these patients, 23 patients met inclusion criteria for PTE and were utilized for the final analysis. Mean age was 38 ± 13.5 yr, and 65% of patients were male (Tables 1 and 2). Mean follow-up time was 73 ± 45.8 mo. The majority (60.9%) of patients in the present cohort had moderate/severe brain injury as defined by Annegers et al.²⁰ Of all patients, 39.1% had a history of mild TBI. Nineteen patients (82.6%) had MTS, 3 had LNC epilepsy (13%), and 1 had NNC epilepsy (4.3%; Figure 1). Eight (34.8%) patients underwent Phase II monitoring. Within the study period, we performed phase II monitoring in a total of 81 patients including those without PTE. Fourteen patients (60.9%) underwent TL, 2 underwent resection of a cortical focus (8.7%), and 7 received VNS (30.4%; Figure 2). The most common reason for VNS implantation was localization to multiple foci (3 patients had bilateral temporal disease and 1 patient had 3 cortical foci). Other reasons for VNS implantation included nonlesional epilepsy (1 patient), refusal of phase II monitoring and resection (1 patient), and Wada test failure in a patient with left MTS (Table 2). Three patients in the VNS group underwent RNS system placement after VNS was unsuccessful in reducing seizure frequency more than 50%.

TABLE 1.

Patient Characteristics – Resection Group

Patient	Sex	Age	Injury	Phase II	Follow-up (mo)	Localization	Procedure	Engel class	Histology
1	M	29	Mild	No	159	Temporal	TL	1	MTS
2	M	23	Moderate/severe	No	83	Temporal	TL	2	MTS
3	M	21	Moderate/severe	No	147	Temporal	TL	1	MTS
4	M	31	Moderate/severe	No	51	Temporal	TL	1	MTS
5	F	57	Moderate/severe	Yes	102	Temporal	TL	3	Gliosis
6	M	44	Moderate/severe	No	91	Temporal	TL	1	MTS
7	F	49	Mild	No	30	Temporal	TL	1	Gliosis
8	F	19	Mild	No	80	Temporal	TL	1	MTS
9	M	41	Mild	No	31	Temporal	TL	1	MTS
10	M	28	Mild	No	59	Temporal	TL	2	MTS
11	M	50	Mild	No	33	Temporal	TL	1	MTS
12	F	67	Moderate/severe	No	49	Temporal	TL	3	MTS
13	F	28	Moderate/severe	Yes	18	Temporal	TL	2	MTS
14	F	36	Moderate/severe	Yes	29	Temporal	TL	1	MTS
15	M	22	Moderate/severe	No	45	Cortical	Cortical resection	1	Gliosis
16	M	40	Moderate/severe	Yes	34	Cortical	Cortical resection	1	Noninfectious cyst

F (female), M (male), MTS (mesial temporal sclerosis), TL (temporal lobectomy).

TABLE 2.

Patient Characteristics – Neuromodulation Group

Patient	Sex	Age	Injury	Phase II	Follow-up (mo)	Localization	Procedure(s)	Contraindication to resection	Seizure reduction % (VNS)	Seizure reduction % (RNS)
17	F	32	Moderate/severe	Yes	56	Cortical	VNS	Nonlesional	0
18	M	52	Mild	No	82	Temporal	VNS	Failed Wada	78
19	M	59	Moderate/severe	No	112	Cortical	VNS	3 foci	33
20	F	29	Mild	No	42	Nonlocalizing	VNS	Patient refused phase II/resection	50
21	M	22	Moderate/severe	Yes	182	Bilateral temporal	VNS, RNS	Bilateral temporal	0	29
22	M	38	Mild	Yes	137	Bilateral temporal	VNS, RNS	Bilateral temporal	33	0
23	M	54	Moderate/severe	Yes	35	Bilateral temporal	VNS, RNS	Bilateral temporal	20	0

F (female), M (male), RNS (responsive neurostimulation), VNS (vagal nerve stimulation).

FIGURE 1.

Distribution of PTE subtypes. The percentage of patients with MTS, LNC epilepsy, and NNC epilepsy is shown here.

Figure 2.

Distribution of treatment modalities. The majority of patients were treated with TL. Other treatments included resection of cortical focus, VNS implantation, or RNS implantation.

Of the 14 patients undergoing TL, tissue pathology was consistent with MTS (hippocampal cell death and gliosis) in the majority (85.7%) of patients. Pathology showed hippocampal gliosis in the remaining 2 patients in this group. One of the 2 patients undergoing resection of a cortical lesion showed cortical gliosis without dysplasia. The remaining patient had a cyst that was resected and was negative for infection.

In the patients undergoing resection of cortical focus or TL (n = 16), 68.8% had Engel class I outcome, 18.8% had Engel class II outcome, and 12.5% had Engel class III outcome (Figure 3, Table 1). No patients in this group were Engel IV after surgery. To examine if PTE patients faired similarly to patients with medically refractory epilepsy treated with TL, we compared our PTE outcomes to historical nontraumatic epilepsy controls and to those of Wiebe et al.¹¹ In our PTE TL cohort, 9 out of 14 (64.3%) were Engel grade I following surgery (Table 1). The mean follow-up time in this group of patients was 70 ± 43 (range 18-162) mo. In our historical cohort, 61% (51 out of 84) were Engel grade I at a mean follow-up of 71 ± 44 (range 13-180) mo. In the Wiebe et al¹¹ study, 58% (23 out of 40) patients in the surgically treated group were Engel I at 1-yr follow-up. There was no significant difference between our outcomes in the PTE cohort and the outcomes in the historical cohort or the Wiebe et al¹¹ cohort (P = .892, Chi-square test; Figure 4). The mean age of our historical cohort was 37 ± 11.6 yr old, and there were 32 males and 58 females in this cohort. There was no significant difference between the PTE cohort and the historical control cohort in age (P = .870, unpaired t-test), sex (P = .146, Fisher's exact test), or length of follow-up (P = .942, unpaired t-test).

FIGURE 3.

Distribution of Engel outcomes in patients treated with resection. The majority of patients were Engel I after surgery. No patients were Engel IV.

FIGURE 4.

Distribution of Engel outcomes in patients who underwent TL. There was no significant difference in outcomes between the PTE patients treated with TL in the current study compared to the outcomes of historical controls or patients in the Wiebe et al¹¹ study.

In the patients undergoing treatment with VNS, average seizure frequency reduction was 30.6% ± 25.6%. Three patients underwent RNS system placement after VNS was unsuccessful in reducing seizure frequency more than 50%. In these patients, seizure severity was decreased but average seizure frequency reduction was only 9.6% ± 13.6% with a mean follow-up of 20 ± 13.1 mo. In the RNS patients, baseline seizure frequency was considered as the seizure frequency after VNS placement since the VNS systems were not removed or turned off prior to RNS implantation.

DISCUSSION

In the present study, we reviewed our institutional experience with the surgical management of PTE. This study adds to the existing, yet sparse, clinical literature on surgical interventions for PTE. In keeping with prior studies, the subtypes of epilepsy were varied within our cohort, although most cases in the present series localized to the temporal lobe, and a majority of these patients had MTS. We found that isolated MTS treated with TL was associated with favorable outcomes (Engel class 1) in most patients, similar to TL outcomes in epilepsy patients without a history of TBI.^11,21

PTE can be broadly categorized topographically into temporal and nontemporal lobe epilepsy; epilepsy phenotypes can further be divided into lesional and nonlesional, with MTS as an additional categorization unique to temporal lobe epilepsy. Temporal lobe epilepsy is the most commonly reported subtype in series of PTE, and MTS is observed in 44% to 53% of these patients.^8,14,22 Unifocal localization is observed most frequently; however, multifocal, bilateral, and poorly localized epilepsy are not uncommon but are inconsistently described and reported.^8,14,17,18 A principle challenge in the surgical treatment of PTE is adequate localization in the presence of multifocal structural abnormalities frequently observed on imaging after head injury.²³ In such cases, invasive monitoring may provide critical data regarding seizure focus localization, although this modality presents unique challenges given that significant adhesions and scarring may preclude safe and effective placement of strips and grids.¹⁸ In this regard, the alternate technique of stereotactic implantation of intracranial depth electrodes largely circumvents this issue. Correlation of structural imaging findings and phase II monitoring findings is not reported consistently in the literature, however, it needs to be evaluated in a rigorous fashion with large, prospective series.

Histopathologic studies in PTE patients have demonstrated that diffuse temporal neocortical and hippocampal cell loss is present in most cases undergoing surgery. This pattern of neuronal loss resembles basic models of TBI, supporting the notion that head injury may precipitate temporal lobe epilepsy.²⁴ This predilection for temporal lobe epileptogenesis is consistently observed in prior published surgical PTE series. Gupta et al,¹⁴ in their series of 22 patients with surgically treated PTE, reported that 73% of patients had temporal lobe epilepsy, and most of these cases were attributable to MTS. In our cohort, pathology showed that majority (85.7%) of patients had MTS while the remaining patients in this group had hippocampal gliosis. The authors found that 76.7% of patients with MTS undergoing surgery were seizure free at follow-up; this is comparable to the rate of 64.3% observed in the present cohort. Similarly, a study by Hartzfeld and colleagues¹⁷ of 53 patients with surgically treated PTE specifically assessed outcomes for the 53% of patients with mesial temporal lobe epilepsy. In comparison to patients with MTS without a traumatic etiology, no statistically significant difference in outcomes was seen, with generally favorable response appreciated (63% vs 78% seizure free at follow-up for PTE and non-PTE, respectively). Our results support these findings, as we observed a seizure freedom rate of 64.3% in the patients treated with TL.

Previous studies have reported that MTS is found in 30% to 35% of PTE patients.^8,22 In our cohort, however, the vast majority (82.6% of patients) had MTS. This discrepancy is likely due to the fact that only patients who received surgery were included in the present study.

Patients in the current study with localization to multiple foci, nonlesional epilepsy, refusal of phase II monitoring and resection, or Wada test failure were offered VNS implantation. Randomized control trials (RCTs) of VNS for medically refractory epilepsy have demonstrated decreased seizure frequency and improvement in quality-of-life.^25,26 Few series have specifically examined VNS placement in PTE patients. In a large, case-controlled study of epilepsy patients undergoing VNS, patients with PTE had greater improvements in seizure control compared to non-PTE patients (73% fewer seizures at 24 mo compared to 57% fewer seizures at 24 mo).²⁷ Moreover, a meta-analysis published by Englot et al²⁸ examined predictors of seizure frequency reduction and found that PTE was associated with greater seizure reduction following VNS when compared with idiopathic epilepsy (78.6% ± 8.7% seizure reduction in the PTE goup). Interestingly, we found an average seizure frequency reduction of only 30.6% ± 25.6% in our cohort. The reason for the reduced efficacy in the current cohort is unclear, but may reflect differences in patient populations.

Newer neuromodulation therapies such as RNS may represent an alternative therapy to VNS for patients that are not surgical candidates for surgical resection. RNS systems work in a closed-loop fashion, where the neural activity of predetermined epileptogenic foci is analyzed for epileptiform activity and a stimulus is delivered to interrupt that activity before it propagates to an ictal event.²⁹ An RCT comparing therapeutic to sham stimulation in patients with refractory epilepsy of all types showed seizure reduction by half over 2 yr (greater than at 1 yr), demonstrating efficacy that increases with the duration of treatment.³⁰ This patient population again was not exclusive to trauma. In our series, 3 patients underwent RNS system placement after VNS was unsuccessful in reducing seizure frequency more than 50%. In these patients, seizure severity was decreased but average seizure frequency reduction was only 9.6% ± 13.6%. Mean follow-up time after implantation of the RNS was 20 ± 13.1 mo, so continued improvement may be seen as the stimulation algorithms are refined.

As in prior reports, we included patients with mild head injury in the present cohort. Indeed, there is evidence within the literature that even mild head injury increases the risk of epilepsy, although not to the same degree as more severe head injury. Christensen et al³¹ identified a relative risk of 2.22 (95% confidence interval [CI] 1.24-1.85) of PTE with mild head injury, although the relative risk (RR) was greater for severe head injury (RR 7.4, 95% CI 6.16-8.89). Similarly, in their series of 5984 patients with TBI, Annegers et al²⁰ determined that mild head injury was associated with 1.5 (95% CI 1.0-2.2) increased risk of epilepsy, while moderate (RR 2.9, 95% CI 1.9-4.1) and severe head injury (RR 17.2, 95% CI 12.3-23.6) were associated with greater risk of epilepsy. A more recent study has demonstrated increased odds of epilepsy in patients with mild TBI (Odds Ratio 1.28, 95% CI 1.07-1.53).³²

Limitations

This study has several limitations. First, the study design is retrospective in nature. It is possible that certain PTE patients did not undergo phase I or II monitoring or surgery because they were thought to be poor surgical candidates, which introduces selection bias. However, surgical lists were tabulated in a prospective fashion and most data collection related to epilepsy history was obtained from comprehensive preoperative epilepsy surgical case conference records, limiting selection and observational bias. Second, there was significant variability in follow-up time, again potentially introducing bias into the outcomes analysis. However, there was no difference in follow-up times between patients with favorable and unfavorable outcome. Finally, the sample size of the present study was relatively small, limiting generalizability of findings, although our results are comparable to other published studies on the surgical management of PTE.

Future studies would benefit from prospective multicenter collaboration to enhance patient sample size and generalizability of findings. With the aforementioned limitations in mind, our study supports phase I/II monitoring in PTE patients followed by resection of an epileptic focus if possible. Our data are most compelling for resection of epileptogenic temporal lobe foci. While our data show that resection of a cortical focus resulted in excellent outcomes, the limited number of subjects in this subgroup make generalization more difficult. Furthermore, future studies investigating the differences between responders and nonresponders of neuromodulatory therapies in PTE patients would help better guide management in this difficult-to-treat patient population. The difficulty of treating this patient population is may be due to the presence of multiple foci, the history of a traumatic etiology, or a combination of both. Future studies could help disambiguate these possibilities.

CONCLUSION

Here, we detail our institutional experience with surgical treatment of PTE patients. We demonstrate that seizure freedom (Engel I) in PTE patients treated with TL was similar to reported descriptions of seizure freedom in non-PTE patients. In our small sample, patients treated with surgical resection of a cortical focus also had high rates of seizure freedom. Patients who were not candidates for resection demonstrated variable response rates to VNS or RNS system implantation. This study supports surgical treatment of patients with medically refractory PTE, especially in patients who are candidates for treatment by resection. Neuromodulatory interventions may play a role in the management of PTE patients with multifocal disease, but further study is necessary to assess the effectiveness of these approaches.

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

COMMENT

The authors present their experience with patients sustaining post-traumatic epilepsy (PTE) submitted to resective or neuromodulatory surgery. PTE occurs in in 10%-25% of patients with moderate to severe injuries and has different subtypes. Most are related to mesial temporal esclerosis (MTE)^1,2 which is corroborated by a recent prospective cohort that patients who evolve with PTE presents at the acute phase with hemorrhage in temporal lobe.³ Although there some data regarding surgical treatment of PTE, the authors present their experience and their different surgical options to manage PTE. As previously found, there was similar benefits of performing temporal lobectomy in those who had MTE, compared to non-traumatic MTE patients.

The vagal nerve stimulator was implanted in just 7 patients, but for PTE, data in literature is even more scarce.^4,5 In this study, the seizure control was not good in patients with VNS but considering the small sample, no conclusions should be made.

Moreover, although the data showed that ressective surgery is an interesting option for refractory traumatic epilepsy, no definitive conclusion should be made due to the weakness of the present design.

Robson Luis Amorim

Sao Paulo, Brazil