The Progression of Epilepsy


Address correspondence to Dr. Warren T. Blume, London Health Sciences Centre, 339 Windermere Road, London, Ontario, Canada N6A 5A5; E-mail:


Summary:  Prognosis for seizure control and cognitive development varies considerably among syndromes. Several factors may interact to influence outcome of an epilepsy including a causative etiology, ictal and interictal discharges, seizure-related trauma or systemic perturbations, and antiepileptic drug (AED) effects. Clinical evidence convincingly supporting Gowers' hypothesis that seizures beget seizures is lacking. Short-term seizure suppression by early treatment does not appear to influence long-term prognosis.

Malignant epilepsy syndromes usually begin in infancy or childhood, have a high seizure frequency, resist the initial AED, and are often associated with progressive cognitive dysfunction. Prompt management of some severe epilepsy syndromes may lessen cognitive decline. However, aggressive AEDs therapy must be balanced against the potential for cognitive side effects, particularly if multiple AEDs are used.

Several experimental paradigms closely parallel human TLE as both have an initial precipitating injury (IPI), a latent period, then recurrent spontaneous seizures. In humans, an IPI is any medical event with neurological implications. Although transition from a latent period to a seizure disorder certainly constitutes “progression” of the disorder, convincing clinical evidence of subsequent worsening has not emerged. Substantial clinical and experimental evidence indicates some cognitive regression and focal atrophy with time for TLE and other intractable syndromes. However, seizure frequency and severity, established early in the disorder, appear stable in most patients, and even regress in benign syndromes. Factors mitigating or extinguishing epilepsies need to be further sought.

Progress of a disorder such as epilepsy can refer to its development or advancement over time (1). As this definition implies, aggravation and the outcome of seizure disorders varies considerably among patients. This article focuses on the course of the epilepsies, including their progression and regression.

In addition to a plan of management, a patient needs some indication as to the course an illness is expected to take. Hence, forecasting a range of outcomes for the individual patient with epilepsy comprises a principal role of the consultant. This article reviews factors that influence the course of a seizure disorder including the seizure tendency, its associated features or both.

There are several aspects of such evolution: clinical, electroencephalographic, cognitive, imaging, and social. This article discusses the first four of these.



Table 1 Lists several factors that conspire to influence the course of epilepsy.

Table 1. Confounding clinical variables in assessing course of epilepsy
filled circle in circleUnderlying disease
filled circle in circleEffect of epileptic discharges on human brain
filled circle in circleNeuropathological effects of seizure-related systemic perturbations and
filled circle in circleAntiepilepsy drug effects including inappropriate drugs
filled circle in circleMultiple drug resistant transporters

An early clinical consideration should be whether an intractable focal epilepsy represents a progressive process such as a tumor, a malformation of cortical development, or an entity such as mesial temporal sclerosis (MTS) that can cause medically intractable seizures (2–4). Semah et al. (5) found a wide range of seizure-free percentages among several etiologies (Table 2). Disorders such as progressive myoclonus epilepsy underlying an intractable generalized epilepsy must be considered (6).

Table 2. Seizure type, etiology, and remission
Seizure typeSF × 1 yr (%)
  1. Source: Semah et al. 1998.

  2. SF, seizure free; TLE, temporal lobe epilepsy; MTS, mesial temporal sclerosis.

Idiopathic generalized82
Cryptogenic focal45
Symptomatic focal35
Extratemporal focal36
TLE without MTS31
Dual pathology 3

Cerebral effect of epileptic discharges

See subsequent sections for review of this topic.

Seizure-related systemic perturbations and trauma

Status epilepticus (SE) may be the only epileptic seizure associated with a subsequent cognitive decline, and this appears to be minor (7). Seizure-related significant head injuries occur in a minority of patients(8,9), but head injury may contribute to encephalopathy in some patients with severe epilepsy involving falls. Pulmonary edema and rhabdomyolysis may damage the brain from anoxia and renal failure, respectively (10).

Antiepileptic drug effects

Meador (11) has found cognitive side effects for monotherapy to be mild to modest for the most commonly used AEDs when the dose remains in the therapeutic range. Similarly, Aldenkamp et al. (12) found motor speed to be the only cognitive function impaired by AEDs in children. Meador's review (11) found no convincing differential cognitive effects among several AEDs with the possible exception of a greater effect by phenobarbital. However, serum levels close to the toxic range and polypharmacy will impair cognition, giving to some a false impression of seizure-related intellectual decline.

Inappropriate choice of AED or too low a dose may also give a false impression of seizure disorder worsening. Some AEDs may increase some seizures in incidence or severity as may occur when drugs for focal epilepsy are used to treat primary generalized seizures (13). Some AEDs may fail to cross the blood–brain barrier due to changes in the expression of multidrug transporter proteins, contributing to AED resistance (14,15).


Epidemiologic data indicate that most epilepsy patients have an excellent prognosis. As the annual incidence for epilepsy among several studies ranges from 20/100,000 to 70/100,000 the prevalence would be 2–5% if no permanent remission occurred whereas actual prevalence hovers around 0.5% (6,16). Nonetheless, seizures in about 30% of patients remain medically refractory (17).

The hypothesis of Gowers

Do seizures beget seizures as Gowers (1881) hypothesized? Hauser and Lee (18) presented a 5-year follow-up of 122 patients considered to have a low recurrence risk after a first seizure (no neurological antecedent, negative family epilepsy history, and no generalized spike-waves on EEG); risk of recurrence increased with each subsequent seizure (25%, 64%, 75%). Although these data could argue for Gowers's hypothesis, it remains probable that the seizure tendency was established prior to the first attack or that certain patients were less able to avoid provocative factors such as sleep loss and stress. Indeed early treatment of seizures, although effectively suppressing seizures in the short or medium term, does not appear to influence long-term seizure prognosis for most syndromes and etiologies (19–26). Exceptions may be infantile spasms and West syndrome, Landau-Kleffner syndrome, and Lennox-Gastaut syndrome (LGS) but early treatment of these may favorably influence cognitive rather than seizure outcome (27).


The epilepsy syndrome may be the best indicator of long-term prognosis for an individual patient (16). Syndromes may be divided according to prognosis for seizure control into malignant, i.e., almost always intractable; variable, i.e., moderate interpatient differences; and benign, i.e., seizure control with monotherapy and possible permanent remission.

Malignant epilepsy syndromes

Cognitive decline can occur in both the generalized and regional malignant epilepsy syndromes. LGS occupies a nosological core among generalized malignant syndromes. An arrest of intellectual development over a 12-year span occurred in 80% of 68 patients with slow spike-waves on EEG, reflecting the LGS in most patients (28). In a different 15–20-year follow-up study of LGS, the proportion of moderately retarded or worse increased from 53% to 93% (29). A hypothesis for the mechanisms of such intellectual deterioration has been proposed (30). Nonetheless, such studies failed to reveal worsening of the seizure component of LGS although a severe seizure disorder persists in at least 50% of patients and even improves in up to 33% (31).

The seizures of Landau-Kleffner syndrome, a regional malignant disorder, gradually regress leaving dysphasia, learning impairment, and behavioral disturbances in about 60% (16).

Clinical identification of syndromes destined for seizure intractability and generalized or regional cognitive impairment involve aspects outlined in Tables 3 and 4. Common features include usual onset in infancy or childhood, high seizure frequency and initial failure of an appropriate AED, EEG with abnormal background activity, diffusely or regionally and/or abundant diffuse or focal epileptiform activity (spikes or spike waves).

Table 3. Generalized malignant syndromes
  1. Compiled from the following sources: (16, 32–38).

  2. SMEI, severe myoclonic epilepsy of infancy; LGS, Lennox-Gastaut syndrome; PME, progressive myoclonus epilepsy.

Ohtahara1–3 monthsBurst suppression
Epileptic spasms (West)3–10 monthsHypsarrhythmia
SMEI1–4 yrNormal, then generalized polyspike-waves
LGS2–5 yrSlow spike-waves
Myoclonic absence1–7 yrPolyspike-waves
PME2 yr–adultPhotoparoxysmal to low flash rates; generalized polyspike-waves
Table 4. Regional malignant syndromes
  1. Compiled from the following sources: (16, 32–38).

  2. HHE, hemiconvulsions hemiplegia epilepsy; CSWS, continuous spike-waves during sleep; MTLE, mesial temporal lobe epilepsy.

HHE2–3 yrRegional spikes, seizures, and delta
Landau-Kleffner (CSWS)3–8 yrTemporal-parietal delta, spike-waves
Rasmussen's encephalitis1–15 yrFrequent widely propagating spikes, seizures
Focal, lesional2–30 yrMany focal spikes, seizures
MTLE1–30 yrTemporal spikes, delta, and seizures

Syndromes of variable prognoses

Table 5 lists these entities. TLE is treated in a subsequent section.

Table 5. Syndromes with variable prognoses
  1. FS, febrile seizures; CEOS, childhood epilepsy with occipital spikes; CAE, childhood absence epilepsy; JAE, juvenile absence epilepsy; JME, juvenile myoclonic epilepsy; SW, spike-waves; PSW, polyspike-waves; PPR, photoparoxysmal response.

Complicated FS6 months–4 yrFocal delta, spikes
CEOS1.5—17 yrOccipital spikes or normal
CAE4–8 yr3 Hz SW, PSW; PPR
JAE10–12 yr4 Hz, SW, PSW; PPR
JME10–15 yr4 Hz SW; PSW; PPR

The EEGs of this group consist of normal background activity and a variable quantity of bisynchronous epileptiform activity, e.g., spike-waves, for generalized entities and regional spikes for focal disorders.

Childhood and juvenile absence epilepsies are a prototype of such epilepsies (39). In a 15-year follow-up study of childhood absence (40), 65% were in remission. Another study found a high correlation between spike-wave quantity on awake “resting” EEG and number of reported absence seizures (41). These children and adolescents generally have normal intellect although a minority have learning and behavioral difficulties (42). No evidence of aggravation of such seizure disorders has emerged from these articles.

Syndromes with excellent prognoses

Table 6 lists syndromes with an almost uniformly favorable outlook. These entities are (by definition) never intractable and usually begin at an older age than do syndromes of a less brighter future. The EEG provides useful guidance; its “background” activity is normal and epileptiform activity is not abundant. A notable exception is the abundant, characteristic spikes of BECT (38).

Table 6. Syndromes with excellent prognoses
  1. BFNC, benign familial neonatal convulsions; BMEI, benign myoclonic epilepsy of infancy; BPEI, benign partial epilepsy of infancy; FS, febrile seizures; BECT, benign epilepsy of childhood with centrotemporal spikes; GMA, “grand mal” on awakening; JME, juvenile myoclonic epilepsy, SW, spike-waves; PSW, polyspike-waves; PPR, photoparoxysmal response.

BMEI4 months–3 yrSW; PSW
BPEI3–20 monthsNormal or focal seizures
FS6 months–4 yrSW; normal
BECT3–13 yr"Rolandic" spikes
GMA6–35 yrNormal or 4 Hz SW; PPR
JME10–15 yr4 Hz SW; PPR

Although mild scholastic problems occur in some children with benign syndromes, the seizure and cognitive outcomes are normal in the vast majority (12,43–46).


Although experimental epilepsy models depict epileptic seizures more than epilepsy syndromes, SE in the immature or mature rat or rabbit, induced by kainic acid, pilocarpine, or electrical stimulation, produces an epileptic condition and pathology with several parallels to the human mesial temporal epilepsy syndrome. Both consist of an initial precipitating injury (IPI), a latent period, then recurrent, spontaneous seizures. Corresponding to the experimentally induced SE, an IPI is clinically any significant medical event with neurological implications occurring prior to a seizure disorder (47). A complex febrile seizure is a common example. Hippocampal sclerosis, the neuropathological signature of mesial temporal epilepsy, is far more commonly preceded by an IPI (87%) than is TLE due to a spontaneously occurring lesion (14%) (47). Before spontaneous seizures supervene, latency periods occur lasting 2–3 months in the rat model and a median of 7 years in the human with considerable variability in each (47–49). Seizure frequency increases over the succeeding several months in most rats but stabilizes in some (48). Increases in apparent human temporal lobe seizure complexity occur in a minority of patients (49) but improved observer or historian descriptive capacity during each patient's seizure duration may have confounded the data. In contrast, video recording of children with TLE disclosed more motor (posturing of limbs, head nodding) phenomena in children less than 6 years old compared to older children whose ictal semiology resembled that of adults (50). This implies less temporal seizure propagation with age. Unfortunately, time to establish failure of a second thoroughly tried AED (51) does not equate to latency of intractability as intractability may have been present long before the current standard of “proof” is attained. I remain unaware of any direct study of human TLE ingravescent time. Confounding factors (Table 1) may alter any estimate considerably.

Nonetheless, clinical and experimental evidence of possible progression of epileptogenesis exists. The transition from the latent period to onset of the seizure disorder constitutes progression in itself. Niedermeyer (52) described an increasingly apparent anterior spike discharge over time in association with TLE, perhaps reflecting greater temporal discharge propagation (53). Experimental production of a secondary “mirror” epileptiform focus that discharges independently of a primary focus has been reported (54). However, were this aggravation inexorable reflecting ever enlargement of epileptogenesis locally, development of additional foci or both; an inverse correlation between temporal lobectomy effectiveness and duration of the disorder would be expected. However, several series have shown duration not to influence outcome (55–58). Taken together these data may reflect an initial ingravescent period followed by one of stabilization at individual levels of seizure disorder severity.

Several structural alterations occur in hippocampal sclerosis that increase excitation. The most characteristic of these is mossy fiber sprouting in the dentate gyrus with extension of axones into the inner molecular layer beginning within one to 2 weeks after kainate SE (48, 59,60). Corresponding with this growth is development of recurrent excitation among granule cells as EPSCs (48). Over several months as recurrent excitatory circuits increase in association with robust axonal sprouting, epileptiform bursts can be elicited when inhibition is suppressed or extracellular potassium is elevated (61,48,62). Such epileptiform bursts have also been described in slices of human hippocampus from patients with intractable TLE (63). Kindling can produce similar mossy fiber sprouting, progressing with the evolution of kindled seizures (64). Mossy fiber sprouting is more likely due to seizure-induced nerve growth factor (NGF) increase than to hilar cell loss (65).

Seizure models applied to immature animals and thus clinical seizure disorders beginning in youth may permanently imprint an immature pattern of seizure development and propagation by preserving normally eliminated pathways (66,67). Presumably, these increased connections would enhance excitability and would contribute to the lowered seizure threshold in adult mammals when seizures occur in development (66,68,69). Preexisting lesions of the immature cortex, while not necessarily producing early seizures, may lower the seizure threshold. For example, a cortical freeze lesion can lower the threshold for hyperthermia-induced convulsions in immature rats (70). Such seizures of the immature brain predispose the animal to greater neuronal injury from adult seizures (71,72). This relationship raises the possibility that a complex febrile seizure-TLE sequence is preceded by preexisting genetic or structural abnormalities in some patients (73).

Decreased inhibition from hilar interneuron loss after SE constitutes an additional factor promoting excitation in these models and in humans (48,67,68). Loss of calcium-binding proteins Calretinin and Parvalbumin in interneurones would have augmented toxic effects of seizure-induced calcium influx (74).

Several factors mitigate these aforementioned excitatory mechanisms. A principal aspect may be the apparent “bidirectional” nature of seizure-induced neuronal loss (see further) and alterations in neural circuits may endow resistance to further seizure-induced change perhaps reflecting early death of most vulnerable cells (48,75.76). Does nature provide its own selective temporal lobectomy?

Some aspects of inhibition may increase in these models. Kainic acid-induced seizures in the developing brain may augment paired-pulse inhibition in the dentate gyrus in adults (62). Although, kainic acid and pilocarpine models of SE may destroy interneurons normally mediating dendritic inhibition, somatic GABAergic inhibition is spared (77). Similarly, glial and neuronal glutamate transporter protein quantity varies among hippocampal regions in SE models. The permanent transporter decrease in the inner molecular layer of the dentate gyrus could contribute to progression of the seizure disorder. However, this effect could be mitigated by transporter increases in somata of CA1-3 neurons and surviving granule cells (78).

Some modulatory effects on hippocampal excitability involve kindling, a model of progressive epileptogenesis. Epileptic seizures increase expression of brain-derived neurotrophic factor (BDNF) in the hippocampus; infusion of BDNF into the hippocampus delays development of kindling and suppresses rapid-kindling-induced seizures (79). Previous kindling may alter the “second hit” effect of kainic acid: fewer kindled rats responded to kainic acid with seizures while those that did respond had some generalized convulsions and faster development of limbic SE in one study (80). Moreover, kindling appeared to provide neuroprotection as hippocampal, piriform, and substantial nigra damage was less than in controls. A similar inhibiting effect of previous kindling on subsequent epileptogenesis was noted by Wada (81). He noted that a kindling-induced secondary antiepileptogenesis can occur at an independent EEG mirror focus to kindled amygdala: An after discharge may be elicited but the contralateral amygdala cannot be kindled.

Finally, heat shock proteins, induced by various insults to the central nervous system, may be neuroprotective in rats (82) and humans (83). Whether their accumulation after prior seizures in adult rats plays a role in the subsequently higher seizure threshold is speculative, but such protection may impede any epileptogenic cascade of pathological events, e.g., by preserving inhibition.

Hippocampal cell loss

SE in kainic acid, pilocarpine, or electrical stimulation models is associated with hippocampal cell loss, particularly in the hilus (48,76,84). Pitkanen et al. (84) examined hippocampal pathology in rats at various intervals after SE evoked by electrical stimulation of the amygdala. They concluded that hilar cell loss correlated positively with duration between SE and sacrifice and not with the number of spontaneous seizures occurring after SE. Gorter et al. (85) reached the same conclusion in a study of parahippocampal neuronal damage in a similar rat model. Similarly (and surprisingly), no correlation was found between mossy fiber sprouting and spontaneous seizure quantities in a chronic in vivo model of Pilocarpine induced SE (86) while a prior study suggested that mossy fiber sprouting was not necessary for spontaneous seizures to occur (87).

Study of hippocampal neuronal densities in resected specimens of TLE patients is consistent with experimental data (47). A temporal lobe seizure duration of about 30 years was necessary before neuronal cell count negatively correlated with duration; for most patients with TLE of about 16 years duration, IPI was principally responsible for hippocampal neuronal loss. The late-occurring negative correlation suggested that recurrent limbic seizures may further damage the hippocampus.


Cross-sectional MRI volumetric studies document hippocampal atrophy ipsilateral to temporal lobe seizure origin as a function of TLE duration (88–91). Bilateral reductions in extratemporal white matter, principally ipsilateral to temporal seizure onset, were also disclosed in the Seidenberg study. In addition to duration, high lifetime seizure number, number of generalized tonic-clonic (GTC) seizures, and a history of IPI, particularly a complex febrile seizure, appeared to contribute to such atrophy in these four studies.

Congruent with the volumetric studies are results of temporal lobe N-acetylaspartate:creatine (NAA/Cr) ratios in TLE patients as compared to controls in a cross-sectional study (92). NAA reflects neuron function and density. NAA/Cr ratios of 60% of their 82 patients were only ipsilaterally lower than controls whereas ratios of the remaining 40% were bilaterally below normal, lower ipsilaterally to seizure origin than contralaterally. Both ipsilateral and contralateral ratios were lower in patients with GTC seizures. Ipsilateral and contralateral ratios correlated somewhat negatively with duration of seizure disorder (R =−0.3 ipsilateral, −0.32 contralateral).

Thus both volumetric and spectroscopic imaging studies are congruent with the neuronal density study all suggesting progressive neuron loss with duration of the disorder but that other factors e.g., IPI and GTC, likely play significant roles.

Cognition and TLE

The most common intellectual deficit associated with TLE is recent memory impairment (93). Verbal memory loss, reflecting an epileptogenic lesion in the language-dominant temporal lobe is the most disabling impediment. However, correlating with generalized as well as hippocampal MRI-disclosed cerebral atrophy in childhood onset TLE, general intellectual status may also be reduced (94). Entorhinal cortex atrophy and dysfunction may also impair memory (95).

Several interrelated factors contribute to these handicaps: any IPI, an epileptogenic lesion, age of onset, duration of the seizure disorder, seizure type and frequency, any seizure-induced cerebral pathology, direct or indirect AED effects and social isolation. Data relating to some of these aspects follow.

The IPI exerts a major impact on memory as disclosed by lower initial memory function than controls (96) and lower hippocampal neuron densities at all epilepsy durations in resection specimens of epilepsy patients compared to autopsied controls (47). A cross-sectional study of memory as a function of duration of disorder disclosed a slope of decline similar to that of the general population (96). Given the lower starting point, memory impairment became clearly more handicapping with time among the TLE patients. Declining intellectual and memory performance in childhood-onset TLE patients is most evident in those with less formal education (94,97). Memory declined in 17% of seizure-free TLE patients, compared with 21% with 2–12 seizures per year, and 38% for more than 12 seizures annually (96). A subsequent longitudinal study (98) also found a progressive cognitive declined among chronic, uncontrolled TLE patients, principally affecting memory. Factors influencing seizure quantity such as IPI severity may also have affected memory decline.


Substantial cognitive and neuroimaging data document decline in these areas for epilepsy syndromes of moderate or severe severity. However, persistent worsening of the associated seizure disorder was not found in those epilepsies whose etiologies are nonprogressive.

The nature and severity of epilepsy pathophysiology becomes established at or soon after seizure onset in most patients with most syndromes. Subsequent variations in severe quantity or severity more likely reflect fluctuations of nonneurophysiological variables such as stress, intercurrent illnesses, management, and compliance. The seizures of patients with all syndromes may improve to various degrees with time and with maturation of the central nervous system, especially those with favorable prognoses. However, physiological and nonphysiological factors are entwined in this circumstance as well. The seizure disorder of a minority of patients with nonprogressive etiologies may gradually and consistently worsen with time but solid evidence of this is lacking.

At or soon after onset, seizure disorders may plateau at levels determined by pathophysiological mechanisms established before clinical manifestations. These include etiology, location, or distribution of the epileptogenic region(s) and the general background of the individual patient. “Progression” would occur during this interval and for various durations thereafter but a plateau is attained. The epilepsy may then stabilize itself by destroying epileptogenic neurons as occurs in temporal lobe epilepsy, or by excitatory processes becoming coupled with inhibitory mechanisms. A search for factors involved in such transitions and how they can be influenced would have therapeutic value.