A study of mortality after temporal lobe epilepsy surgery

Statistical analysis.

Univariate logistic regression analysis was used to examine the relationship between side of surgery and postoperative seizure outcome in the overall group and in each of the pathologic subgroups. For the purpose of analysis, patients with an Engel class 1A/1B outcome were considered seizure free and compared with other outcome classes combined (classes 1C/1D/2/3/4). The differences between side of surgery and class 1C/1D outcome were also analyzed for the mesial temporal sclerosis (MTS) group. Mortality rates per person-years of follow-up in the left and right MTS groups were compared using an approximate normal test for equality of rates.13 A p value < 0.05 was considered significant.

Epilepsy-related deaths and relation to postoperative seizure control.

Table 3 lists the details of each patient and the circumstances of death. Thirteen deaths were considered epilepsy related, not related to operative complications: SUDEP (definite and probable) (6), status epilepticus (2), drowning in the bath (2), aspiration during a seizure (2), and accident (1). Unsubstantiated status epilepticus was recorded as cause of death on the death certificates of three patients who were reclassified as definite or probable SUDEP. Only 3 of the 13 patients were seizure free before death; 2 of these were classified as SUDEP and the third was killed by a passing motor vehicle while in an apparent confusional state. One patient (Patient 12, table 3) had been seizure free for 4 years and then experienced generalized tonic-clonic seizures (GTCS) at a rate of one every 3 months. Death was due to drowning in the bath while 5 weeks pregnant. The postmortem carbamazepine levels were 2 mmol/L (quoted therapeutic range, 4 to 12 mmol/L). Another patient (Patient 16, table 3) had been seizure free for 2 years before regular complex partial seizures (CPS) with secondary generalization recurred. Death occurred 1 month after insertion of a vagal nerve stimulator. Six additional patients (nos. 2, 3, 6, 7, 8, and 11) had at least a 50% reduction in seizures, but five of these were noted to have more severe or generalized seizures. Only one patient continued to have seizures of the same kind and frequency as before surgery, and this patient died from status epilepticus (Patient 9).

Mortality and the effect of side of surgery and operative pathology.

Of the 13 late deaths attributed to epilepsy, 11 occurred in patients who had MTS as operative pathology. Although the mortality rate was highest in patients with nonspecific findings at pathology, this is misleading in view of the few person-years and the inclusion of all deaths due to early and late operative complications in this group, with only one death in this group considered seizure related. In the group with other lesions, one of the two deaths was possibly epilepsy related, although it is uncertain whether the terminal event occurred in the context of a seizure (Patient 14, table 3). A significant difference was observed between deaths in patients with left and right MTS. Patients with right-sided MTS experienced a postoperative mortality of 1 in 53 years (SMR 32.0), whereas those with left MTS had a mortality of 1 in 310 years (SMR 3.3); p = 0.0128. Although more SUDEP deaths occurred in patients with right MTS compared with left (4 versus 1), death also occurred from status epilepticus, seizure-related drowning, seizure-related aspiration of vomitus, accident, and suicide. Only two epilepsy-related deaths occurred in the left MTS group despite 930 years of follow-up compared with 533 years of follow-up in patients with right MTS.

Previous reports of mortality after epilepsy surgery.

A number of reports4,14-19 have addressed the mortality rates of patients undergoing epilepsy surgery. These are summarized in table 4. Most conclude that surgery leads to a reduction in mortality. Little information is given as to exact cause of death or details of the clinical course after surgery. Comparison with the present series can only be approximate because mortality per person-years of follow-up has not been ascertained in the larger studies. Furthermore, reflecting differences in the patient populations, two of the reports16,17 describe many deaths occurring as a result of progression of temporal neoplastic disease. This was not encountered in our series of surgery for the sole purpose of alleviating seizures and not for the treatment of underlying mass lesions. The high rate of suicide in the Danish series (Jensen14) (36% of deaths) and early Maudsley series (Taylor,15 Bruton16) (22% of deaths) is in contrast to our findings of only one case (5% of total). This may be in part related to the operated population, at least in the Maudsley series (a major psychiatric institution from which many patients in the series of Falconer20 were referred). However, it is also possible that there has been a real decrease in the incidence of postoperative suicide in recent years.

Previous studies of mortality in patients with epilepsy.

Relationship of mortality to postoperative seizure outcome.

Although surgery for TLE reduces mortality, it is clear that mortality remains significantly raised. It seems intuitive to assume that this continued mortality excess is simply explained by those patients whose seizures are not influenced by surgery. However, as detailed in table 3, most of the patients who died, including the deaths from seizure-related causes, had considerable benefit from surgery in terms of seizure relief. Seven patients were seizure free, or 41% of the group that died (excluding operative complications), which is not significantly different from the absolute seizure freedom rate for the group that survived (40%). Two other patients had prolonged seizure-free intervals of 2 and 4 years, and five patients had at least a 75% reduction in postoperative seizure (usually CPS) frequency.

Little attention has been paid to the frequency and nature of postoperative seizures in the previous studies of mortality after epilepsy surgery. Yet it is recognized that in up to 50% of nonsurgery patients who die suddenly seizures are infrequent,28 and 32% have fewer than three per year.29 In a study of SUDEP cases identified through self-referred relatives, 25% had had less than 10 GTCS in a lifetime.30 Jensen14 refers to one postoperative patient who was seizure free for 3 years and who reportedly died from status epilepticus after antiepileptic drug (AED) withdrawal. Only one other report has highlighted the possibility of sudden death after apparently successful anterior temporal lobectomy in one patient who was seizure free for 9 months and in another who had a 75% reduction in seizures.31 The majority of patients with SUDEP have a history of GTCS. At least seven of the patients who experienced epilepsy deaths in the present study had substantial relief of their preoperative CPS but continued to experience at an increased frequency or developed de novo tonic-clonic seizures. The occurrence of infrequent, often nocturnal, generalized seizures after TLE surgery has been recognized before,32 but rarely quantified in the literature. In a single study, Blume and Girvin33 reported that, although generalized seizures were reduced from 70 to 39% after lobectomy, 23% of patients who had not had GTCS preoperatively had at least one after lobectomy. The relatively infrequent GTCS in their study were related to suboptimal AED treatment. Our results suggest that this altered seizure pattern may be an important mortality risk factor after surgery. The patient details in table 3 emphasize that this phenomenon may be important not just for SUDEP but for other seizure-related deaths such as status epilepticus and drowning and supports the continuation of some AED therapy long-term, particularly in patients who have a tendency to generalized seizures on attempted AED withdrawal. The question of AED compliance has been raised in relation to sudden death in epilepsy.34,35 In our series, no firm data were available to implicate this as a major causative factor, although Patients 10 and 12 (see table 3) had subtherapeutic levels of carbamazepine at autopsy.

Relationship of mortality to operative pathology and side of resection.

Little information is available on the relationship between operative pathology and mortality in the previous reports of mortality. In the series of Falconer and Serafetinides20 no seizure-related deaths occurred in patients with foreign tissue lesions at pathology, and all epilepsy-related deaths occurred in patients with either hippocampal sclerosis or nonspecific findings.16 In our series, almost all seizure-related deaths occurred in the MTS group. This may indicate that patients with this lesion are particularly liable to have generalized seizures postoperatively despite significant alleviation of preoperative partial seizures. A striking finding of this study was the extremely high mortality rate in patients who had right-sided resections for MTS. This patient subgroup had an SMR of 32.0 and a mortality rate of 1 per 53 person-years, a rate that exceeds any so far reported for cohorts with intractable epilepsy. In patients with right MTS, 9 of the 10 deaths were epilepsy related, and the other was a suicide in the context of postoperative depression. The percentage of patients achieving a class 1A/1B outcome in right MTS was only 28% compared with 48.5% of patients with left MTS (p = 0.013). Because all epilepsy-related mechanisms of death were represented in the right MTS group, the lower seizure remission rate was likely the most significant factor in the high mortality rate of this group. Furthermore, with regard to seizure severity, right MTS patients had more frequent class 1C/1D outcomes (infrequent disabling seizures or seizures occurring only in relation to changes in AEDs) compared with left-sided patients, although the difference was not significantly different (see table 1). We are unable to explore this further because we do not have accurate information concerning the occurrence of GTCS at any time during the average of 9.13 years of postsurgical survival of each patient in the whole cohort (average clinical follow-up for MTS patients was 4.5 years).

The significant difference in seizure outcome between left and right in our series only applied to the group with MTS and not to the group as a whole. Previous reports of outcome after TLE surgery have not indicated that right-sided resections have a worse seizure outcome. The contribution of resection side to outcome after surgery has been addressed in relatively few previous reports.36-42 No definite conclusions can be drawn from these reports. In all of these series (except that of Falconer and Serafetinides20), the pathology of resection could only be determined for a proportion of cases and, even where pathologic information is complete, no study has analyzed the effect of resection side by pathologic subgroup. The poor outcome for right MTS in this present study may be due to chance, but a critical review of the literature forces the conclusion that this issue has not been reviewed systematically until now.

There is no definite explanation as to why right-sided resections for MTS have fared badly. Results from an ongoing study42 of preoperative predictors of outcome after temporal lobectomy in our series have not demonstrated a significant difference in the frequencies of preoperative variables such as perinatal insults, complicated febrile seizures, family history of epilepsy, psychiatric history, duration of seizure disorder, age at operation, secondarily generalized seizures, and interictal temporal EEG activity between patients who had left and right MTS. These variables have been assessed using actuarial statistics with regard to their influence on probability of achieving 12 consecutive months of postoperative seizure freedom. The only significant predictor of outcome that differed between the two groups was the presence of a prior brain injury (either perinatal damage or infectious or traumatic injury excluding febrile seizure), with only 33% of patients with right MTS, who had this risk factor, achieving 12 months of seizure freedom compared with 69% of cases without this risk factor (p = 0.033, log-rank test). No such difference applied to patients with left MTS. Subjects with right-sided hippocampal sclerosis may have more widespread cerebral abnormalities that account for their persistent seizures and increased mortality after surgery. In support of this, another study,43 using quantitative post-processing of preoperative MRIs, has described widespread extra-hippocampal abnormalities in patients with hippocampal sclerosis and persistent seizures after surgery. These findings were more common (p < 0.05) in patients who had had right-sided operations.

The present findings support previous suggestions that overall mortality and epilepsy-related mortality are reduced after TLE surgery and add further strength to the medicoeconomic and quality-of-life issues pertaining to epilepsy surgery. However, in the subgroup with right-sided MTS, mortality remains similar to groups with intractable epilepsy and to patient cohorts who have been evaluated for surgery but not operated upon. We suggest that infrequent, either new or continuing, convulsive seizures postoperatively may be a major risk factor for this excess mortality. Although discontinuation of AEDs after surgery may be considered as the ideal operative outcome, particularly in children, we suggest that when generalized seizures occur postoperatively, even if infrequent, adequate AED drug treatment should be recommended. In that situation, satisfactory seizure control may be reliably maintained with monotherapy.44 Prospective multicenter studies are required to replicate our findings, particularly with regard to the differential effects on seizure control and mortality of left- and right-sided resections for TLE.