How Long Do Most Seizures Last? A Systematic Comparison of Seizures Recorded in the Epilepsy Monitoring Unit


Address correspondence and reprint requests to Dr. S. Jenssen. Hahnemann University Hospital, Broad and vine Streets. Mail Stop 308, Philadelphia, PA 19102. E-mail:,sigi


Summary: Purpose: More information is needed regarding how long seizures typically last, since this influences treatment decisions. Seizure type and other factors could influence seizure duration.

Methods: Data were collected from a random sample of patients being evaluated with continuous video and scalp EEG. Seizure duration was defined as time from early sign of seizure (clinical or EEG) until the end of seizure on EEG. Seizures were categorized as simple partial (SPS), complex partial (CPS), secondarily generalized tonic–clonic (SGTCS), primary generalized tonic–clonic (PGTCS) and tonic (TS). SGTCS were divided into a complex partial part (SGTCS/CP) and a tonic–clonic part (SGTCS/TC). Median and longest duration of each seizure type in each individual were used. Comparisons of seizure types, first and last seizure, area of onset, and state of onset were performed.

Results: Five hundred seventy-nine seizures were recorded in 159 adult patients. Seizures with partial onset spreading to both hemispheres had the longest duration. SGTCS were unlikely to last more than 660 s, CPS more than 600 s, and SPS more than 240 s. PGTCS and TS had shorter durations, but the number of subjects with those two types was small. CPS did not differ in duration according to sleep state at onset nor side of origin.

Conclusion: A working definition of status epilepticus in adults with cryptogenic or symptomatic epilepsy can be drawn from these data for purposes of future epidemiologic research. More information is needed for the idiopathic epilepsies and in children.

Epileptic seizures are discrete paroxysmal electrical events, typically ending after a brief period of time. In some cases, however, seizures can become persistent and continuous, presumably because the mechanisms that allow seizure activity to end have failed. This constitutes status epilepticus (1). Unfortunately, there is no consensus as to when the usual seizure duration has been surpassed and the condition of status epilepticus has been reached. A better knowledge of the typical duration in different seizure types would help physician determine when status epilepticus is present. We therefore set out to measure the duration of seizures in hospitalized patients using video-EEG monitoring.


Data from patients being evaluated with continuous video and scalp EEG in the Epilepsy Monitoring Unit at Thomas Jefferson University Hospital, Philadelphia, Pennsylvania, were reviewed retrospectively from 2000 to 2002. The patients had been admitted for diagnostic evaluation or for emergent treatment of uncontrolled seizures. The vast majority had either cryptogenic or symptomatic epilepsy. A 10–20 electrode placement system with additional sphenoidal electrodes was used in a 32 channel acquisition system. Most patients were examined during seizure activity.

Anticonvulsant medications were lowered or increased according to clinical indications. Since medications often were gradually tapered during the hospital stay, the seizures recorded later in the sequence for a given patient often occurred while the patients were on lower medication doses than for earlier seizures. We cannot determine the precise extent to which this is true because drug levels were not monitored during the hospital stay.

To be included all seizures had to have some ictal EEG change. We defined beginning of the seizure as the first EEG change or first subjective or objective clinical change whichever occurred first. The end of the seizure was defined as the end of ictal EEG activity. We divided seizures into the following categories: simple partial seizure (SPS), complex partial seizure (CPS), secondarily generalized tonic–clonic seizure (SGTCS), primary generalized tonic–clonic seizures (PGTCS), and tonic seizures (TS). SGTCS were subdivided into a first complex partial portion (CP/SGTCS) and second tonic–clonic portion (TC/SGTCS). Seizures without clinical change and only ictal EEG activity were categorized as SPS if ictal change appeared in only one hemisphere or as CPS if the ictal EEG changes were bilateral. Basic descriptive statistics were obtained on the five major seizure types studied (SPS, CPS, SGTCS, PGTCS, and TS) for the longest seizure per patient, the median seizure per patient, and the range of durations per patient. These seizure types were then compared to each other on those dependent variables using Kruskal–Wallis ANOVAs, followed up by Bonferroni-corrected Mann–Whitney U tests (p-value < 0.005 was required for significance).

CPS were divided into several overlapping subcategories and their durations were compared in the following manner: first seizure was compared to last seizure in subjects with more than one, and first seizure with onset from wakefulness was compared to first with onset from sleep in patients with both, using the Wilcoxon signed-rank test (because these were within-patient comparisons). Subjects with CPS from the right temporal lobe were compared to subjects with CPS from the left temporal lobe using the Mann–Whitney U test (because these were comparisons between different groups of patients who were consistent in onset localization).

The duration of each seizure type was correlated with the number of seizures recorded for each type using the Spearman test.


Table 1 shows the descriptive statistics on the median seizure duration per patient, longest per patient, range of durations per patient, and actual number of recorded seizures of that type per patient. Because the data are nonnormal, we utilized medians and ranges, rather than mean and SD, for summary purposes. Kruskal–Wallis ANOVAs comparing the seizure types on median duration per patient (p < 0.001), longest per patient (p < 0.001), and range of durations per patient (p < 0.01) were significant.

Table 1. Seizure duration statistics by type of seizure (s)
Seizure type (No. of patients)Median duration per patientLongest duration per patientRange of durations within patientNo. of seizures recorded
  1. Data shown are median for each parameter (picking either the seizure of median duration for each patients or the seizure of longest duration in each patient). Ns for range are smaller than values shown since patients with only one seizure did not have a range. Seizure types are ordered by the median value of the longest duration per patient. All seizure durations are in seconds. See text for abbreviations.

TS(7)  18.5(8–410)34 (11–620)21.25 57
SPS(25) 28(3–180)35(3–475)23.00 65
PGTCS(6) 66(59–75)68(59–76) 2.50  8
CPS(85) 78(8–298)11 (40–960)46.00375
SGTCS(34)130 (37–139)142 (50–630)37.00 70

Multiple comparison tests (Bonferroni-corrected Mann–Whitney U) on median duration per patient showed that the seizure types differed significantly from each other. SGTCS lasted longer than CPS and CPS and PGTCS lasted longer than SPS. On the other hand, CPS did not differ from PGTCS. TS did not differ from SPS or from PGTCS. Each of the nonsignificant comparisons involves at least one of the types with N < 10.

Multiple comparisons on longest seizure per patient found that SGTCS lasted longer than CPS which lasted longer than SPS (same as for median durations). The same nonsignificant comparisons were found and SPS did not differ from PGTCS. Comparing the range of seizure duration, the only significant post hoc difference was that CPS had a greater range than PGTCS.

For CPS, we assessed whether there were any differences in duration between first and last seizures if multiple seizures had occurred, or whether sleep state (awake vs. asleep) influenced seizure duration. Table 2 shows the medians and ranges for each of these measures. There was a nonsignificant trend toward last CPS to be longer than first (two-tailed p = 0.074 by Wilcoxon test) and no difference in seizure duration between those occurring while awake and those that happened while asleep (p = 0.795). We also compared median and longest per patient CPS durations for left- and right-sided seizures. Table 3 shows the descriptive statistics for these subtypes. Neither comparison was significant (p > 0.15 by Mann–Whitney U test).

Table 2. Comparisons within CPS seizures (s)
CharacteristicMedian durationRangeN
  1. Medians and ranges for first and last CPS and awake and asleep each of these measures. There was a trend toward last CPS to be longer than first and no difference between awake and asleep. All seizure durations are in seconds. See text for abbreviations.

First CPS6723–51771
Last CPS85 8–96071
Awake CPS7223–51719
Asleep CPS9023–47019
Table 3. Left- versus right-sided CPS seizures (s)
  1. Table 3 compares the duration of CPS originating form the left or right hemisphere using the median duration and the longest duration in each patient. There was no difference. All seizure durations are in seconds. See text for abbreviations.

Median duration per patient
 Left 8040–28329
 Right 9029–27028
Longest duration per patient

To compare partial onset with generalized onset TC seizures, we measured the median duration of just the TC component of the SGTCS (TC/SGTCS) and compared them to the median total duration of PGTCS seizures using the Mann–Whitney test. The median duration of TC/SGTCS was 74 s and for the tonic–clonic component of PGTCS was 66 s, which was not different. There was no correlation between duration of each seizure type and the number of seizures recorded for each type.

Since the data include variable numbers of seizures within patients, it is difficult to develop simple guidelines for duration. We chose to focus on the longest seizure per patient, which should thus help to identify upper limits. Fig. 2 displays the seizures of longest duration for each patient in each seizure type. It can help illustrate which durations could be considered the maximal limit for usual duration and when there is a risk of developing status epilepticus:

Figure 2.

Distribution of the single longest seizure duration in each patient by seizure type is shown. The longest per patient is a direct indication of what constitutes extreme seizure duration. We left out for the sake of the figure representation two SGTCS lasting 79 and 41 min, respectively, two CPS lasting 52 min and 16 min, respectively, and one SPS lasting 48 min. For the types with the longest durations, notably SGTCS and CPS, seizures over 7 min (420 s) appear to stand out. SPS of more than 8 min stand out. TS is even more bimodal, with a single very long seizure lasting over 10 min but all the rest of the longest per patient seizures being around a minute or less. See text for abbreviations.

  • 1Of SGTCS only two of the 70 seizures in 34 patients lasted longer than 10 min. One patient had only one SGTCS and this lasted 79 min and another had only one SGTCS which lasted 41 min. These two are left out of Figs. 1 and 2 in order not to expand the vertical axis.
  • 2One out of 85 patients had three CPS that lasted more than 10 min (out of 378 CPS in total). They were 52, 48, and 16 min, respectively. These three seizures are left out in Fig. 2 in order not to expand the vertical axis. This patient had a total of 12 CPS and the median duration was only 36 s.
  • 3One of the 67 SPS in 25 patients lasted 48 min (this patient had three SPS with a median duration of 50 s). Another patient had SPS of 475, 250, and 220 s, but a median duration of 51 s from eight SPS. Otherwise none of the patients had longest duration of more than 200 s.
  • 4Of 10 PGTCS in six patients, none lasted more than 120 s, and also no TC part of SGTCS lasted longer than 120 s.
  • 5Of 52 TS in seven patients, one lasted 620 s. (The same patient had only one other TS and this lasted 200 s.) All other patients had durations of less than 200.
Figure 1.

The distribution of the median seizure duration for each patient of each seizure type is shown and an illustration of what is most common is provided. We left out for the sake of the figure representation two SGTCS lasting 79 and 41 min (only seizures). In the figure most median SGTCS durations are less than 300 s and all median CPS durations are less than 300 s while all median SPS durations are 180 s or less. PGTCS vary little and TS has a bimodal distribution. See text for abbreviations.


This analysis provides further evidence that seizure duration differs for the various seizure types. This is the first study that systematically compares the duration of different seizure types. For seizures with a focal onset, duration increases with increasing seizure severity and extent of spread. Hence, SGTCS last longer than CPS which last longer than SPS. This discrepancy may be related to activation of neuronal networks involved at the onset and during propagation of seizures. There is evidence that the EEG during a seizure transitions toward an increased electrical signal complexity preceding seizure termination (1). These dynamic changes might take place earlier when a smaller part of the brain is involved. Seizure duration did not follow a normal distribution and some seizures of extreme duration were recorded, suggesting a breakdown of normal seizure termination mechanisms.

In animal models several factors may affect seizure duration. These include loss of inhibitory influences from thalamus or other subcortical structures (2), impaired GABA transmission (3), changes in adenosine receptors (4,7), pH (5), endogenous cannabinoids (6), intracellular calcium (7), or ketosis secondary to low calorie diet (8). The TC phase may also provide a forceful impetus for seizure termination. This could happen through increased GABA transmission (9) or through the suppression of reentry circuits (10). There is anecdotal evidence that electroconvulsive therapy can terminate status epilepticus (11) presumably through a similar mechanism. Magnet activation of a vagus nerve stimulator can help terminate a seizure probably exerting its effect through activation of brainstem structures (12). There is also evidence that inheritance can promote the tendency of prolonged seizures; mechanisms are as yet unknown (13).

Our data on the duration of TCS resemble that of previous reports on hospitalized patients. A meticulous analysis of the TC phase of SGTCS (equivalent to the TC part of SGTCS in our study) in 47 subjects found a mean duration of 67 s (similar to our mean of 74 s) and concluded that SGTCS seldom last longer than 2 min (14). Another report found that duration of the tonic phase of generalized TCS was 20 s and clonic phase lasted approximately 30 s (15). A third study that included 23 patients with 50 SGTCS found a mean total seizure time of 109 s and a mean time of the TC part of 53 s based on video and EEG recordings (16).

Four of the 579 seizures we recorded lasted more than 30 min. This is similar to what has been reported previously on the incidence of seizures of more than 30 min duration in the video and EEG monitoring unit (18).

Finally, another study reported longer durations of seizures, but differs both in methods and population studied (17). Seizure duration was recorded based on witness reports only, and all seizures were new onset seizures in children who later developed epilepsy. This study found that unprovoked seizures lasting more than 10 min and even 30 min are common in children. The longer durations in this study as compared to our data could be because children typically have seizures of longer duration, because new onset seizures typically have a longer duration or because the duration was overestimated by the witnesses. This needs further investigation.

In patients who had more than one CPS, the last seizure tended to last longer than the first. If this finding can be replicated, it could be because of decreased anticonvulsant medications during the hospital stay (19). Drug levels were not obtained on each day during hospitalization, so seizure duration could not be correlated to medication changes. We also do not have data to correlate seizure duration with time to clinical recovery to baseline mental state.

We did not find that the subject's state of consciousness at the onset of the seizure affected the duration in CPS, nor did the duration of CPS with left versus right onset differ. While it might be posited that synchronization that takes place during sleep alters seizure duration, this was not supported by the data. Known differences in hemispheric anatomy and function also do not appear to be related to seizure duration.

The limitations of this report arise from the homogeneous population being studied, with nearly all patients having intractable epilepsy and a majority of these being cryptogenic partial epilepsy with frequently a long duration of epilepsy. Our population included only a small minority of patients in late childhood and no infants so the data can not be generalized to children and infants (16). A correlation of seizure duration in individual patients with previous history of status epilepticus and, in some of the cases, subsequent clinical outcome of epilepsy surgery, could be attempted in the future (20).

From a practical point of view, our data can help offer an empirical definition of status epilepticus and an indication for when to intervene by giving medication and instituting urgent care. Based on the data on longest seizure in each patient, we propose that

  • 1SGTCS longer than 660 s (11 min) are likely to extend into status epilepticus.
  • 2CPS longer than 600 s (10 min) are likely to extend into status epilepticus.
  • 3SPS longer than 240 s (6 min) are likely to extend into status epilepticus.

These three recommendations are based on 25 or more patients and 67 or more seizures of each type. For TS, while we observed 52 seizures, these were in only seven patients. Durations of over 200 s seem long for TS, but this needs confirmation. PGTCS were all fairly short with no outliers and no seizures lasted longer than 120 s (2 min) but we had only 10 seizures in six patients, which makes a strong conclusion impossible.

In summary, though seizures in adult epilepsy vary widely in duration, they do differ consistently between seizure types. More data are needed in children and idiopathic epilepsy.