The commission wishes to acknowledge the support of UCB Pharma for logistical support in the organization of this workshop.
Address correspondence and reprint requests to Michael R. Sperling at Thomas Jefferson University Hospital, 900 Walnut St., Suite 200, Philadelphia, PA 19107, U.S.A. E-mail: email@example.com
Summary: Mortality rates are higher in people with refractory epilepsy than in the general population. We assessed mortality rates in a prospectively followed cohort who had epilepsy surgery, to examine the factors related to mortality and to assess the relationship between seizure control and mortality. Five hundred eighty-three patients were evaluated. Mortality was strongly related to seizure control (p = 0.001), with 18 deaths observed in patients with recurrent seizures (mortality rate = 11.4 per 1,000 person-years) and 1 death in patients with no recurrent seizures (mortality rate = 0.85 deaths per 1,000 person-years). Patients with generalized epilepsy who had corpus callosotomy had a higher mortality rate than patients who had resective or transective surgery. The side of surgery and gender did not influence mortality rates. The standardized mortality ratio was 5.75 for patients with recurrent seizures and was significantly higher for females than males. These data show that the excess mortality associated with refractory epilepsy is eliminated after epilepsy surgery when seizures are abolished and suggest that epilepsy surgery reduces the risk of epilepsy-associated death.
A variety of limitations and handicaps are imposed on people who suffer from epilepsy, of which the associated psychosocial disabilities receive much attention. However, serious medical complications can occur, and the most disturbing is the diminished life expectancy in those who harbor the condition (1–10). The increased risk of dying is not uniform among patients with epilepsy, for several factors affect mortality. The chief influences appear to be etiology, seizure type, associated mental handicap, seizure control, and perhaps duration of epilepsy (1,2,6,11–13). There is a greater risk of death in those with remote symptomatic epilepsy than idiopathic epilepsy (5,6,13), and tonic–clonic seizures probably pose greater risk than other types (1). The coincidence of mental retardation and epilepsy markedly raises mortality rates (1), as does lack of seizure control (1,2,8,10). Some studies also suggest that the risk of death is greater in the early years after developing epilepsy than afterwards (4,6,13).
Two papers have recently been published regarding mortality after epilepsy surgery (2,14). The patients in these studies are unique in the medical literature, for they not only have medically refractory epilepsy, but their diagnosis has been incontrovertibly established by video-EEG monitoring. The subjects in these cohorts had similar interventions, permitting the examination of the relationship between different diagnoses and the success of the intervention with subsequent mortality. We previously reported postoperative mortality in 393 patients after epilepsy surgery in the United States (2). Approximately half had no postoperative seizures, and their mortality rates were indistinguishable from that of the general population. In contrast, individuals who had any recurrent seizures after surgery had a markedly elevated mortality rate, with a standardized mortality ratio (SMR) of 4.69 and a risk of death of 13.7 per 1,000 person-years. We concluded that uncontrolled seizures are a major risk factor for excess mortality in people with refractory epilepsy. Furthermore, we suggested that epilepsy surgery reduces the risk of dying in patients with uncontrolled seizures, since elimination of seizures reduced the mortality rate in patients who became seizure-free. Subsequently, Hennessy and colleagues reported long-term mortality rates in 305 patients from the United Kingdom. They found an SMR of 4.5 for their entire cohort, with a mortality rate of 7.4 per 1,000 person-years. They found that SMR was higher in patients who had right-sided mesial temporal sclerosis than left-sided disease, but these patients had higher seizure recurrence rates than patients with left-sided disease.
The present study was undertaken to reassess mortality in a larger cohort of patients with refractory epilepsy after epilepsy surgery. The primary aim was to further examine the relationship between seizure control and mortality after surgery. Secondary measures included assessing the relationship between mortality and other variables including laterality of resection, gender, and type of operation. With a larger patient sample of nearly 600 patients, we also hoped to derive more accurate estimates of mortality rates and SMR for the subpopulations within the cohort. We tested the hypotheses that mortality was related to seizure control after surgery, and that it was related to laterality and gender. We also wished to determine if the SMR reverted to normal in patients whose seizures stopped after surgery.
All patients who had epilepsy surgery between February 1, 1986, and June 30, 2000, were included in this review. The censoring date used for last possible follow-up was September 15, 2000. All patients had a neurosurgical procedure for treatment of refractory seizures, including resection, multiple subpial transection, and partial or complete corpus callosum section. Individuals with malignant brain tumors were excluded. All patients had a multidisciplinary evaluation, including history and physical examination, long-term video-EEG monitoring, magnetic resonance imaging, neuropsychological testing, Wada test, positron emission tomography (selected patients), and intracranial EEG monitoring (selected patients). Resection and multiple subpial transection were performed for partial epilepsies with a localized epileptogenic zone. Anterior or complete corpus callosotomy was performed for idiopathic or symptomatic generalized epilepsy, typically for uncontrolled tonic, atonic, or tonic–clonic seizures.
Preoperative and postoperative data were recorded for each patient in a prospective database containing complete preoperative and postoperative clinical data. Information relevant to this study included gender, age at surgery, type of operation, date of operation, presence or absence of postoperative seizures and date of the first postoperative seizure if one occurred, postoperative seizure frequency, date of last contact, and date and cause of death. Sudden unexplained death in epilepsy (SUDEP), including probable sudden unexplained death, was registered as the cause of death using established criteria (15). Patients were followed from the time of surgery to date of death or date of last contact. All postoperative seizures were registered as a recurrence except for atypical focal motor seizures that occurred in the first week after surgery (1 patient); these seizures were excluded since they reflect transient postoperative inflammation and are unrelated to long-term seizure outcome. All other seizures, including those related to medication withdrawal, were counted as recurrences. Patients with only auras after surgery were included in the seizure-free group.
Mortality was evaluated in the cohort as a whole and in subgroups. Patients were classified into two groups depending on the type of surgery. Patients who had any resection or subpial resection were grouped together for analyses since they had partial epilepsy. These patients were further subdivided as either having had an anterior temporal lobectomy or extratemporal operations in some analyses. Patients who exclusively had a partial or complete corpus callosotomy were grouped together since they had generalized epilepsy. The mortality rates were described in terms of deaths per 1,000 person-years for the population as a whole and then for various subgroups. The groups mentioned above were separated into two categories: those with any recurrent seizures after surgery and those with no recurrent seizures. All follow-ups commenced at surgery. For patients who recurred, all postoperative follow-up time was registered as recurrent time, and all follow-up time for patients who did not recur was registered as nonrecurrent time.
The SMR provides a means of comparing the mortality rate in a particular subject population with that of the general population. Therefore, in comparing SMRs between different groups of patients, one can derive an estimate of how they fare in comparison with their age- and gender-matched peers, which is different than comparing how their mortality rates directly compare with one another. SMRs were calculated by comparing the number of observed deaths with the number of expected deaths among an age- and sex-matched group of U.S. residents not racially or regionally divided (16). An expected probability of death was calculated for each subject and summed over each member of the comparison groups to provide group-specific estimates of the expected number of deaths; 95% confidence intervals were constructed around the SMR (17).
The Kaplan–Meier method was utilized to assess survival, using the Cox–Mantel test to compare groups. Analyses were performed assessing survival for the entire cohort, and for subsets as mentioned below. The main comparisons were between patients with recurrent seizures and patients with no recurrent seizures, though subsidiary analyses compared survival for patients with different types of operations, different gender, and different laterality of surgery. All analyses were performed with Statistica (Statsoft, Tulsa, OK).
A total of 583 patients were evaluated in this study, which includes the 393 patients reported in our previous analysis (2). Three hundred eighteen were males (30.9%) and 265 were females (69.1%); 521 (89.4%) had resection or subpial transection, and 62 (10.6%) had a callosal section. The mean age of surgery was 33.3 ± 10.3 years (range 10.7–66.2 years) and average follow-up duration was 4.9 ± 3.2 years. The average full scale IQ for 535 patients with known IQ was 87.5 ± 14.5, 89.2 ± 13.4 for the resective patients, and 72.2 ± 14.9 for the callosotomy patients.
Survival of cohort
Of 583 patients, 564 (96.7%) survived until study termination, and 19 (3.3%) died. The timing of death is shown in Fig. 1 and causes of death are listed in Table 1. The majority of these patients had a significant reduction in seizure frequency after surgery, and some had had rare seizures. For the cohort as a whole, the mortality rate was 6.9 per 1,000 person-years; 325 patients had one or more recurrent seizures (1,580 person-years follow-up), and 258 patients did not recur (1,180 person-years follow-up). Patients with recurrent seizures were more likely to die in the follow-up period than patients with no recurrent seizures (p = 0.001). Eighteen of the 325 patients who had a recurrence died (5.5%) and 1 of 258 patients with no recurrences died (0.4%) during the follow-up period (Fig. 2). Among patients with seizure recurrence, the mortality rate was 11.4 per 1,000 person-years (or 10.8 per 1,000 person-years excluding the 1 perioperative death). Patients with no recurrent seizures had a mortality rate of 0.85 deaths per 1,000 person-years.
Table 1. Causes of death after epilepsy surgery
Cancer (outside central nervous system)
Motor vehicle accident
Survival of patient subsets
For the 521 patients who had either resective surgery or subpial transection, 265 had recurrent seizures, and 256 had no recurrent seizures. Fig. 3 illustrates survival in these two groups. Twelve deaths were observed among patients with recurrent seizures (4.5%), while 1 death (breast cancer) was observed among patients with no recurrent seizures (0.4%, p = 0.005). The mortality rate for resective or transective patients with recurrent seizures was 9.13 per 1,000 person-years, and it was 0.85 per 1,000 person-years for patients who had no recurrence. The improved survival after surgery was true for patients who had temporal lobectomy (n = 446) as well (p = 0.004), since all deaths after resective or transective surgery occurred among those who had temporal lobectomy.
Patients with generalized epilepsy with anterior or complete callosal sections (n = 62) had a higher mortality rate than patients who had partial epilepsy with focal resection or subpial transection (n = 521) (p = 0.001). However, since all but 3 patients who had a callosotomy had recurrent seizures after surgery, this comparison could be misleading. Therefore, we compared survival in callosal section patients who recurred (n = 60) with survival in resective or transective patients who recurred (n = 265). There was a trend for increased mortality in patients who had callosal section compared with patients who had resection or subpial transection (p = 0.06), as shown in Fig. 4. The mortality rate for patients with callosal section was 22.6 per 1,000 person-years (18.8 per 1,000 person-years excluding the single perioperative death).
Patients who had right-sided operations had the same mortality rate as patients with left-sided operations. There was no effect of gender on mortality rates, although the lower mortality rates in women in the U.S. population yielded an effect on the SMR as described below.
Standardized mortality ratio
The SMR for the entire cohort was more than threefold greater than expected (Table 2).
Table 2. Standardized mortality ratios for patients with epilepsy surgery
Recurrent resection/subpial transection
Recurrent corpus callosotomy
Patients with recurrent seizures had a high SMR (SMR = 5.75, 95% CI 3.51–9.27). Females with recurrent seizures had a higher SMR (SMR = 13.13, 95% CI 6.89–24.17), than males (SMR = 3.06, 95% CI 1.34–6.59).
These results confirm and extend previous findings that survival after epilepsy surgery is strongly related to postoperative seizure control. Of 19 deaths, 18 occurred in patients who had one or more recurrent seizures, despite reduction in seizure frequency in many patients. Patients with recurrent seizures had nearly a sixfold increased number of deaths compared to the expected number. This SMR is quite similar to that reported by Nashef et al. (11) for a population of patients with refractory epilepsy seen in a tertiary referral center.
In contrast, patients who stopped having seizures after surgery had an SMR indistinguishable from that of the general population, consistent with the hypothesis that successful epilepsy surgery reduces the mortality rate. Without surgery, these individuals may have continued to experience seizures and may have had a high mortality rate. This rate would be similar to that observed in refractory patients treated in tertiary care epilepsy centers and in patients in the present series who had recurrent seizures after surgery. Hennessy et al. (14) also observed a lower mortality rate in a cohort of patients after epilepsy surgery than expected in chronic epilepsy, suggestive that surgery reduces mortality. Vickrey et al. (18) also observed a lower mortality in patients treated surgically than those who were not offered surgery, but their nonsurgical group was deemed unsuitable for surgery and therefore did not constitute an adequate control group. A recent study noted no reduction in mortality when surgical patients were compared with a control group, but the patients in this sample were accrued from 1949 to 1988 before modern diagnostic techniques were available, and had high perioperative morbidity and relatively low rates of achieving seizure freedom (19). Nonetheless, these reports are consistent with the findings in the present study and provide further support for the beneficial effect of epilepsy surgery on mortality. We cannot exclude the possibility that underlying biological differences between patients who respond to surgery and those who do not could influence mortality. However, it is unlikely that biological differences explain all of the variance in postoperative mortality rates; the effect of seizure control cannot be disregarded. Studies of people whose epilepsy is medically treated also suggest that well-controlled patients have lower mortality rates than patients with recurrent seizures (10). These findings are entirely consistent with mortality in surgically treated patients.
It is striking that the SMR was substantially higher for females. However, this should not be interpreted that females with refractory epilepsy are more likely to die than males. Males and females had similar mortality rates, and the difference in SMR is an artifact of the SMR method. Since females have a lower mortality rate than males in the general U.S. population, the denominator used in calculating the SMR was considerably smaller in females than males, leading to the consequence of a higher SMR for females. The SMR is also affected by other factors, notably age. A young population with epilepsy has a low death rate and excess deaths from epilepsy are quite noticeable. In contrast, the expected death rate is high in elderly individuals, and their increased mortality from epilepsy might not be terribly conspicuous. Therefore, the real difference in mortality depends on an excess of death in the young. Studies of the effect of duration of epilepsy on mortality are also biased by the SMR method. Since the expected death rate rises with age in adults, the SMR might decrease with the passage of time solely because of the progressively increasing death rate in older individuals.
We found no effect of laterality of epilepsy surgery on SMR, in contrast to Hennessy et al. (14). Their patients with right-sided surgery had a worse seizure outcome than their left-sided patients, possibly explaining their result.
As noted in other studies (3), SUDEP accounted for approximately half the fatalities in the present study. The mechanisms responsible for SUDEP have yet to be elucidated and could be related to either an interictal or ictal disturbance. The present study suggests that seizures play an important role in mediating the pathophysiologic process; this might be inferred from low mortality rates in patients who are controlled by medication.
Lastly, the interaction between seizure frequency and mortality is not known. Is there a safe seizure frequency that does not increase the risk of dying? The extant literature does not provide the answer, but it and the present study suggest that patients with relatively infrequent seizures remain at greater risk for dying (2,3,10). In the absence of conclusive data, it seems safest to assume that a complete lack of seizures is the only satisfactory level of seizure control.
The high mortality rate in people with refractory epilepsy (20) and its apparent reversibility with seizure control warrant an aggressive approach to seizure control, whether through medication or surgery. Future studies are needed to define the pathophysiology underlying the excess mortality and specific risk factors that operate in addition to seizure control.