- Top of page
Summary: Purpose: Whether cerebral damage results from epileptic seizures remains a contentious issue. We report on the first longitudinal community-based quantitative magnetic resonance imaging (MRI) study to investigate the effect of seizures on the hippocampus, cerebellum, and neocortex.
Methods: One hundred seventy-nine patients with epilepsy (66 temporal lobe epilepsy, 51 extratemporal partial epilepsy, and 62 generalized epilepsy) and 90 control subjects underwent two MRI brain scans 3.5 years apart. Automated and manual measurement techniques identified changes in global and regional brain volumes and hippocampal T2 relaxation times.
Results: Baseline hippocampal volumes were significantly reduced in patients with temporal lobe epilepsy and could be attributed to an antecedent neurologic insult. Rates of hippocampal, cerebral, and cerebellar atrophy were not syndrome specific and were similar in control and patient groups. Global and regional brain atrophy was determined primarily by age. A prior neurologic insult was associated with reduced hippocampal and cerebellar volumes and an increased rate of cerebellar atrophy. Significant atrophy of the hippocampus, neocortex, or cerebellum occurred in 17% of patients compared with 6.7% of control subjects. Patients with and without significant volume reduction were comparable in terms of seizure frequency, antiepileptic drug (AED) use, and epilepsy duration, with no identifiable risk factors for the development of atrophy.
Conclusions: Overt structural cerebral damage is not an inevitable consequence of epileptic seizures. In general, brain volume reduction in epilepsy is the cumulative effect of an initial precipitating injury and age-related cerebral atrophy. Significant atrophy developed in individual patients, particularly those with temporal lobe and generalized epilepsy. Longer periods of observation may detect more subtle effects of seizures.
Chronic intractable epilepsy is associated with significant structural alterations both within and beyond the epileptogenic zone. The commonest histologic finding in human intractable temporal lobe epilepsy (TLE) is hippocampal sclerosis (HS), found in 50–70% of specimens obtained from temporal lobe resections (1,2).
Intractable TLE also may be associated with extralesional volume deficits, including the ipsilateral and contralateral temporal lobe; contralateral hippocampus; ipsilateral amygdala, entorhinal and perirhinal cortex, thalamus and caudate, and cingulate gyrus (3–7). Cross-sectional magnetic resonance imaging (MRI) studies of patients with intractable epilepsy also have reported significant cerebellar (8) and cerebral volume reduction (9). Some authors have proposed that extrahippocampal abnormalities may be capable of epileptogenesis because widespread structural deficits and cerebellar volume reduction in mesial TLE have been found to predict poor seizure outcome after temporal lobectomy (8,10).
Despite the prevalence of widespread structural abnormalities in patients with chronic epilepsy, the timing and pathogenesis of such changes remain obscure. The crucial question relates to whether the abnormalities are progressive and the cumulative effect of years of epilepsy (11); the effect of an initial precipitating insult (IPI) during a vulnerable phase of cerebral development (12); the presence of a preexisting developmental abnormality predisposing to insults and further cerebral damage (13); or a combination of these factors acting synergistically (14).
Recent longitudinal studies attempting to address this question have produced disparate results. In a study of 24 patients with newly diagnosed cryptogenic TLE, a strong correlation was found between the number of convulsive seizures and ipsilateral hippocampal volume (HV) loss over a period of 3.5 years (15). These findings were supported by a study demonstrating that the frequency of partial but not generalized seizures was correlated with ipsilateral HV loss in 12 patients with refractory TLE and unilateral HS (16). Although these hospital series suggest that HS and seizure-related hippocampal damage may evolve in the absence of an initial insult [e.g., status epilepticus (SE)] (17,18), a recent longitudinal MRI study comparing eight patients with well-controlled epilepsy with five patients with intractable partial epilepsy found no relation between seizures and either progressive or new structural hippocampal damage (19). Furthermore, group comparisons of patients with newly diagnosed partial epilepsy and age-matched controls found no differences in mean baseline HV and change in HV over a period of 1 and 3.5 years, respectively (20,21). Community-based studies of patients with newly diagnosed epilepsy have shown no effect of epilepsy syndrome, seizure recurrence or antiepileptic drug (AED) use on HV change over a 3-year period (21,22), although in 13% (22) and 8% (20) of individuals in these studies, significant hippocampal atrophy developed.
Serial imaging studies provide an opportunity for monitoring disease progression in vivo. In this study, we aimed to investigate the effect of epileptic seizures on hippocampal, cerebral grey and white matter, and cerebellar volumes, by performing a population-based longitudinal MRI study of patients with epilepsy and age-matched controls. By incorporating image registration and a range of manual, semiautomatic, and automatic segmentation techniques, we compared volume change in patients with TLE, extratemporal partial epilepsy, and generalised epilepsy and determined whether rates of cerebral damage over a 3.5-year period varied according to epilepsy syndrome. Clinical risk factors for the development of significant cerebral atrophy were investigated.
In contrast to hospital-based studies, in this community-based study, seizures were largely well controlled with medication. Only a small proportion of patients proceeded to epilepsy surgery; therefore our findings pertain largely to nonsurgical patients with epilepsy.
- Top of page
Two key questions are addressed in this longitudinal MRI study: (a) whether epilepsy is associated with MRI-detectable cerebral damage over a 3.5-year period, and (b) whether particular clinical risk factors increase individual susceptibility to damage. The main findings were that (a) patients with TLE had significantly reduced HVs and longer HCT2 times than controls, patients with extratemporal partial epilepsy, and generalized epilepsy at baseline; (b) loss of hippocampal and neocortical volume was comparable between patient groups and controls and therefore not syndrome dependent; (c) a history of an IPI was associated with an overall reduction in hippocampal and cerebellar volume, and an increased rate of cerebellar atrophy; (d) convulsive seizures were correlated with change in HV, TBV, and GMV in patients with extratemporal partial epilepsy; (e) in a greater proportion of individuals with epilepsy significant hippocampal and neocortical volume loss developed compared with that in control subjects.
Methodologic issues and limitations of the study
As a population-based study that included patients with newly diagnosed and chronic epilepsies, our patient groups were inevitably heterogeneous, reflecting a diverse range of aetiologies with variable susceptibility to damage. This diversity may have influenced our ability to detect group differences between the epilepsy syndromes but allowed us to explore a wider range of individual variation and is more relevant to the population developing epilepsy than would that in a selected clinic population. Our individual analysis identified significant volume and signal change in a greater number of patients with epilepsy than in control subjects. Although volume changes were more commonly observed in individuals with TLE and generalised epilepsy, this finding was not statistically significant.
Our study is based on the assumption that brain damage is detectable by using a serial MRI approach. Although prospective MRI studies have demonstrated the rapid development of hippocampal damage over several months (18,36,37), it is possible that seizure-induced damage is subtle and not readily picked up by current imaging techniques. Coregistration of serial scans allowed us to improve the variability of repeated measurements and thus reduce change due to measurement error. Some evidence indicates that pathologically proven HS may exist in the absence of a significantly reduced HV (38) or increased HCT2 (32). Consequently, our stringent definition of HS may underestimate the development of HS in patients undergoing either significant HV reduction or HCT2 increase alone. The proportions of patients in whom HS develops de novo may also vary according to whether HS is defined on quantitative MRI or visual criteria, although visual comparison in this study did not observe the development of HS in any patient.
Hippocampal damage in temporal lobe epilepsy
Cross-sectional studies have shown significant ipsilateral HV reduction in patients with frequent seizures and prolonged duration of TLE (39). It has thus been proposed that volume reductions reflect either (a) the cumulative neurobiologic effect of repeated seizures or (b) a bias toward the accumulation of refractory cases after a severe precipitating injury. Our study confirmed that patients with TLE had significantly reduced HVs at baseline. Analysis of longitudinal data revealed similar rates of hippocampal volume reduction in patients with TLE, extratemporal focal epilepsy, and control subjects. No correlation was observed with numbers of convulsive or partial seizures. Patients with histories of neurologic insults showed significant overall reductions in HV, suggesting that for the majority of TLE patients, structural damage is likely to be related to an IPI, with subsequent volume loss being influenced by age rather than by seizures.
In our study, subject groups demonstrated differential changes in HCT2 relaxometry. Patients with TLE and generalised epilepsy demonstrated a nonsignificant increase in HCT2. Studies incorporating absolute cell counts have shown that an increased HCT2 signal reflects the severity of dentate gliosis, with many astrocytes showing glial fibrillary acidic protein (GFAP)-positive properties indicative of recently occurring or ongoing abnormal processes (33). In our patients with TLE, change in HCT2 was not correlated with seizure frequency, AED exposure, or drug intoxication. The reason for the observed decrease in HCT2 relaxometry in control subjects and patients with extratemporal partial epilepsy is not clear. Phantom-based quality assurance data of repeated HCT2 measures at our institution have shown no consistent long-term drift, and subjects were aligned to the same external landmark on each scan occasion. It is possible that HCT2 values may decrease with age with a disproportionate loss of glial to neuron cells in the dentate gyrus.
In contrast to a recent follow-up study of 12 patients with HS (15), in none of the 14 patients with HS at baseline in our study did significant HV loss or increase in HCT2 relaxation time develop over the period studied. Only a relatively low number of patients had HS at baseline, and this is likely to reflect both the broad spectrum of disease severity seen in population-based studies and the exclusion of children with HS seen with seizures before age 14 years. The lack of detectable disease progression in patients with HS could be a manifestation of the “floor effect,” in which the initial insult may have been so severe as to damage the hippocampus to such an extent that no further damage could be observed over the study period (40). In accordance with Salmenperä's study (22), we did not observe the development of HS (using our morphometric criteria) in any patient with a normal scan at baseline.
Although progressive hippocampal atrophy occurred infrequently among our epilepsy population and was influenced primarily by the age of the patient; a complementary voxel-based neocortical interrogation of the same study population (41) has shown that TLE is associated with the development of subtle neocortical atrophy. Temporal lobe atrophy was observed in 17% of patients with chronic TLE and HS, suggesting that neuronal damage may extend beyond the hippocampus to involve the ipsilateral temporal lobe, particularly in the presence of severe hippocampal damage from an initial brain injury.
Cerebellar and neocortical damage in epilepsy
Cerebellar atrophy in epilepsy has previously been considered a consequence of either prolonged PHT therapy (42), seizure-mediated cellular damage through cerebrocerebellar diaschisis (43), hypoxic damage (44), or SE (45). In this study, patients with epilepsy had a reduced cerebellar volume at baseline compared with controls, although this was not significant. Our observation that cerebellar volume loss over time was comparable in patients and controls suggests that these mechanisms cannot wholly explain the volume loss observed in patients with epilepsy, and that at least part of the CBV loss is likely to occur independent of a history of seizures or AED therapy. The association of an IPI with an overall reduction in CBV and increased cerebellar atrophy suggests a putative link between cerebral insults and epilepsy-related cerebellar atrophy. This is consistent with suggestions that cerebellar damage may occur after brain trauma (46), and that loss of inhibitory function in patients with a structurally damaged cerebellum may worsen prognosis for good seizure control (8). Although patients with epilepsy had reduced total brain volumes at baseline and accelerated rates of cerebral atrophy, particularly patients with TLE (Fig. 1D), these findings were not statistically significant.
Initial precipitating insults
A major insult can cause ipsilateral hippocampal damage and damage to other structures, particularly on the side of the seizure focus (36). In their clinicopathologic study (47), Mathern et al. showed that IPIs were important in the pathogenesis of HS, and that HS occurred in patients with extratemporal partial epilepsy only if a prior IPI had occurred. In TLE patients, HS was strongly associated with IPIs involving seizures. Our results showed that a history of an IPI was associated with a significant reduction in hippocampal volume. This observation persisted after exclusion of patients with HS, suggesting that the reduced brain volumes did not simply reflect the greater number of patients with HS in those with a history of IPIs. A significant increase in the rate of cerebellar atrophy was seen in patients with a history of IPIs, which could not be attributed to other potential confounders (e.g., alcohol consumption, age, AED use, or a history of recurrent head injuries between the two scans).
Our results corroborate previous suggestions that, in addition to causing measurable volume changes, an insult may prime the brain, making it more vulnerable to the effect of seizures (48). Experimental studies in rats demonstrated progressive cortical and subcortical neuronal loss for ≤1 year after traumatic brain injury (49,50) and suggested that a chronically progressive degenerative process may be initiated by the injury. Putative mechanisms for the progressive tissue loss observed after brain injury include the consequences of the primary insult (i.e., wallerian degeneration) and progressive secondary injury mechanisms including apoptotic cell death, inflammation,and excitotoxicity in white matter tracts. Consistent with this hypothesis is the finding by Kim and colleagues (51) that those with temporal lobe complex partial seizures associated with an overt structural lesion show less neuronal loss than do those with a history of complicated febrile seizures or an IPI.
Regression analyses showed that only patients with extratemporal partial epilepsy had a significant correlation between frequency of convulsive seizures and change in HV, TBV, and GMV. No correlation was found between seizure frequency and ipsilateral HV loss in cryptogenic TLE, suggesting that regional volume loss seen in association with convulsive seizures is not necessarily localized to the site of seizure generation but may be remote from the epileptic focus. Methodologic differences that may explain the discrepancy between our results and Briellmann's (15) finding that numbers of generalized tonic–clonic seizures are inversely correlated with ipsilateral HV loss, include differences in the populations studied, scanner consistency, blinding procedures, and use of age-matched controls (52).
Despite the prospective documentation of seizures, an accurate seizure count, particularly of focal seizures, was not always attainable because patients were not always aware of their seizures. Nonetheless, it would seem implausible that inaccuracy of seizure recall should affect our observations significantly, because the change in HV in the four groups was strikingly comparable (Fig. 1A). Mathern and colleagues (47) showed that longer durations of TLE were associated with decreased neuron densities in all hippocampal subfields, an observation independent of IPI-induced neuronal loss, although a long time course (>30 years) was required to demonstrate the negative correlation. Thus the authors proposed that the substantial hippocampal neuronal loss observed in HS was likely to be the result of an IPI rather than the effect of repeated limbic seizures. Their suggestion that limbic seizures slowly “damaged” the brain over several decades may contribute to the lack of correlation observed between seizures and hippocampal atrophy in our study.
None of the four patients with SE between the two scans experienced significant brain volume losses or HCT2 changes over the 3.5-year period. Opinion regarding cerebral damage after SE is divided. Although a number of case reports described the development of HS after an acute process (37), Salmenperä and colleagues (40) showed that progressive HV reduction was not an invariable consequence of promptly treated SE.
A previous cross-sectional study showed that men with TLE demonstrated greater brain atrophy compared with women with TLE. Because the number of convulsive seizures contributed significantly to these abnormalities in men but not in women, the authors postulated that men were more vulnerable to seizure-induced brain volume loss, although the initial damage was likely to be gender independent (53). In our study, we found no gender effect with regard to either initial volume loss or ongoing susceptibility to brain damage. Changes were comparable in men and women for all MRI parameters studied.
Exposure to antiepileptic drugs
Our data did not provide evidence for increased cerebral damage with prolonged exposure to AEDs. Although patients with TLE had been exposed to significantly more AEDs than had patients with extratemporal partial epilepsy and generalized epilepsy, the rate of volume loss was comparable in the three patient groups. A recent study comparing HV changes in newly diagnosed patients treated with carbamazepine, vigabatrin, and tiagabine monotherapy found no difference after 3 years of follow-up (22).
The role of PHT in the pathogenesis of cerebellar atrophy has not been resolved in previous retrospective studies. PHT treatment is frequently compounded by the cumulative effect of hypoxia in the context of repeated convulsive seizures. In the present study, we found no association between treatment with PHT—either long-term use or acute intoxication—and whole cerebellar volume. It is possible that seizures or PHT use might exert a selective regional effect, and a more detailed study of the cerebellar hemispheres and lobules of the vermis may be warranted.
The observation that hippocampal and cerebral atrophy in individual patients is not always related to high seizure numbers, in this and other studies (15), implies that factors other than overt seizures have a significant role. In our study, the predominant factors influencing volume loss were a history of an antecedent neurologic insult and age. We therefore believe that it is important to control for these variables when assessing the contribution of seizures on the brain. Although our findings did not support a syndrome-specific effect for regional volume loss, the patient population in our community-based study was heterogeneous, and follow-up of more homogeneous patient groups over a longer time might identify particular characteristics, such as genotype, that place individuals at an increased risk of epilepsy-related damage.
Current efforts to increase reproducibility and reduce operator intervention by automation [for example, by automatic propagation of manually drawn baseline segmentation to coregistered follow-up scans (54)] may provide more sensitive measures in the future, although their current performance does not surpass that of the methods used here. The serial use of new MRI contrasts such as magnetization transfer ratio imaging (55) may identify more subtle changes. Serial MR spectroscopic studies also may be of value in assessing functional changes that are not detected on structural imaging.
In summary, our longitudinal study of 179 patients with epilepsy showed that progressive regional or global cerebral damage is not an inevitable consequence of epileptic seizures. The significantly reduced baseline HV observed in patients with TLE was attributed largely to an antecedent neurologic insult. Subsequent hippocampal and neocortical atrophy was determined primarily by age and was, in the majority of cases, independent of a diagnosis of epilepsy. In a greater number of patients (23% of TLE patients), significantly more volume loss developed over a period of 3.5 years, than it did control subjects. This appeared independent of seizure frequency and AED use and may have occurred in response to an underlying epileptic process. A suggestion exists that an early neurologic insult may prime the brain and enhance age-related cerebellar atrophy. Further studies are required to characterise these neurologic insults because measures such as the prompt treatment of prolonged febrile seizures may modulate the impact of such insults on the brain and reduce subsequent atrophy.