OUP user menu

Affective aggression in patients with temporal lobe epilepsy
A quantitative MRI study of the amygdala

L. Tebartz van Elst, F. G. Woermann, L. Lemieux, P. J. Thompson, M. R. Trimble
DOI: http://dx.doi.org/10.1093/brain/123.2.234 234-243 First published online: 1 February 2000


Recurrent episodes with interictal affective aggression are a rare but well-recognized problem in patients with temporal lobe epilepsy. They are referred to as episodic dyscontrol or, more precisely, as intermittent explosive disorder (IED). The amygdala play a crucial role in the affective evaluation of multimodal sensory input and the neurobiological mediation of aggressive behaviour. With hippocampal sclerosis, in the context of mesial temporal lobe sclerosis, being the most common cause of temporal lobe epilepsy, we hypothesized that the amygdala might be affected by the same pathogenic process in aggressive patients. We investigated 50 patients with temporal lobe epilepsy: 25 with and 25 without a history of IED. Data from clinical, electrophysiological, neuropsychological and psychometric investigations were obtained, as well as MRI scans for the quantitative assessment of possible amygdala pathology. We found no evidence of a higher prevalence of amygdala sclerosis in the aggressive patients. Hippocampal sclerosis was significantly less common in patients with temporal lobe epilepsy and IED. However, a significant subgroup of patients (20%) with temporal lobe epilepsy and aggressive behaviour had severe amygdala atrophy in the context of a history of encephalitis. Another subgroup of aggressive patients (28%) had different left temporal lesions affecting either the amygdala or periamygdaloid structures. IED was associated with left-sided or bilateral EEG and MRI abnormalities, low IQ and high scores in depression and anxiety.

  • amygdala
  • aggression
  • intermittent explosive disorder
  • MRI
  • temporal lobe epilepsy
  • AT2 = amygdala T2
  • IED = intermittent explosive disorder


Phenomenology and classification of aggression

Human aggression is an important social and clinical problem (Fenwick, 1986; Saver et al., 1996; Trimble, 1996; Swartz et al., 1998). The relationship between temporolimbic epilepsy and aggressive behaviour is a particularly controversial issue (Geschwind, 1975). While a history of complex partial seizures is reported to be common in patients with episodic affective aggression (Bach-Y-Rita et al., 1971; Elliott, 1982), most of the community-based studies did not find an increased prevalence of aggressive behaviour in patients with epilepsy (Kligman and Goldberg, 1975; Lishman, 1998).

One problem with studying aggression is its phenomenological and probably neurobiological heterogeneity, leading to difficulties in assessment and classification. However, human data, in agreement with animal research, point to the existence of at least two different phenomenological and neurobiological subtypes of aggression: predatory and defensive aggression (Goldstein, 1974; Mungas, 1983; Moyer, 1987; Vitiello and Stoff, 1997). Predatory aggression is characterized phenomenologically as a well-structured and goal-directed behaviour performed in an emotionally calm and concentrated state of mind. In contrast, defensive aggression is seen typically in the context of high emotional arousal associated with vocalizations and signs of fear or anger. The behaviour itself is less structured and defensive (Valzelli, 1981). Apart from the planned and goal-directed aggression often seen in persons with antisocial personality disorder (predatory aggression), most forms of human aggression are thought to be a reaction towards a perceived threat, be it real or not (Albert et al., 1993). Obviously, the perception as to whether a stimulus is threatening or not is decisive in the information processing leading to the aggressive behaviour.

Neurobiology of aggression

Various brain structures have been implicated in the mediation of aggressive behaviour in animals and humans, the most important being the periaqueductal grey (Brandao et al., 1994; Behbehani, 1995), the hypothalamus (Andy and Jurko, 1972), the amygdala and associated limbic structures (Dicks et al., 1968; Halgren, 1992; Kling and Brothers, 1992; Rolls, 1992; Aggleton, 1993) and the frontal lobes (Damasio et al., 1990; Miller et al., 1997; Raine et al., 1998). Within this network of critical structures, the amygdala are thought to play a crucial role in the mediation of fear-induced aggression, a subtype of defensive aggression (Aggleton, 1993; LeDoux, 1995; Charney and Deutch, 1996; Gallagher and Chiba, 1996). They receive extensive input from various levels of sensory information processing and project to most of the other critical brain structures, i.e. the brainstem, hypothalamus, thalamus and frontal lobe (Amaral et al., 1992; Alheid et al., 1995). From a neurophysiological point of view, they are in a predestined position for the affective evaluation of multimodal sensory input. Thus pathology within the circuits affecting the amygdala might lead to mental states where the misinterpretation of sensory input as threatening leads to aggressive outbursts. In agreement with this assumption, electrical stimulation of the amygdala can lead to experiences such as fear, anxiety or anger (Chapman et al., 1954; Gloor et al., 1982), and lesioning of the amygdala severely impairs fear conditioning in animals (Davis et al., 1994) and humans (LaBar et al., 1995). Furthermore, in an open retrospective study of 481 cases of bilateral amygdalotomies performed for the control of conservatively untreatable aggressiveness, moderate to excellent improvement of aggressive behaviour was reported in 70–76% of cases (Ramamurthi, 1988).

Aggression and temporal lobe epilepsy

The prevalence of aggression in epilepsy in general, not taking into account the specific epileptic subsyndrome, varies between 4.8% (Rodin, 1973) and 50% (Gastaut et al., 1955). In a large survey of 666 patients with temporal lobe epilepsy, Currie and co-workers reported aggression in 7% of the patients (Currie et al., 1971). Falconer reviewed 100 patients from London's Maudsley Hospital referred for temporal lobectomy and found a prevalence of outbursts of aggressive behaviour in as many as 27% of the patients (Falconer, 1973). However, these studies were hampered by selection bias, and the real prevalence of aggressive behaviour in epilepsy remains controversial (Lishman, 1998).

In epilepsy, three different types of aggressive behaviours should be distinguished on the basis of their relationship to the seizures: ictal, post-ictal and interictal aggression. All three types of aggression are defensive, in that they are episodes of affective aggression, i.e. they occur in the context of high emotional arousal, anger or fear.

Ictal and post-ictal aggression are often associated with confusion or psychosis (Treiman, 1991). Interictal aggression, on the other hand, can be seen in the context of an antisocial personality disorder which, in turn, might be the consequence of the sometimes difficult psychosocial background of patients with epilepsy. However, an interictal syndrome of episodic affective aggression, independent of observable ictal activity, major psychiatric disorder or antisocial personality disorder, is well described and has been referred to as episodic dyscontrol (Bach-Y-Rita et al., 1971; Maletzky, 1973; Ratner and Shapiro, 1979; Leicester, 1982; Elliott, 1984; Stone et al., 1986). Episodic dyscontrol is characterized by several discrete episodes of failure to resist aggressive impulses that result in serious assaultive acts or destruction of property. The behaviour is out of proportion to any precipitating psychosocial stressor and is not due to substance abuse, another mental disorder such as personality disorder, any other first axis psychiatric disorder or a general medical condition such as head trauma or neurodegenerative diseases. The phenomenological criteria are those of intermittent explosive disorder (IED) according to DSM-IV (Elliott, 1984; American Psychiatric Association, 1994). Because of the emotional arousal typically seen in episodic dyscontrol, this behavioural syndrome has to be classified as a subtype of defensive aggression. The high level of arousal with signs of anxiety or fear is also the reason why amygdala pathology is thought to contribute to this behavioural syndrome (Fenwick, 1986; Elliott, 1992; Trimble et al., 1996). In three past studies, the relationship between different psychobiological factors, e.g. brain pathology, IQ and demographical background, and aggression in epilepsy was addressed. While Rodin found more evidence of organic brain disease (Rodin, 1973), and Falconer reported an increased incidence of mesial temporal lobe sclerosis in aggressive patients with temporal lobe epilepsy (Falconer, 1973), Herzberg and Fenwick did not find any relationship between specific EEG or CT pathology and aggression in patients with temporal lobe epilepsy (Herzberg and Fenwick, 1988). All three studies found a relationship between low IQ and aggression, and two reported an association between male sex and aggression (Rodin, 1973; Falconer, 1973).

To our knowledge, there is no study investigating a possible association between aggressive behaviour and brain pathology using the modern technique of MRI and focusing on the amygdala as an important limbic brain structure for this condition.

MRI assessment of temporal lobe structures

The most common pathology underlying temporal lobe epilepsy is hippocampal sclerosis often in the context of mesial temporal sclerosis (Margerison and Corsellis, 1966; Wieser, 1983; Gloor, 1991). Some authors give a prevalence to hippocampal sclerosis in temporal lobe epilepsy as high as 65% (Babb and Brown, 1987). A radiological in vivo diagnosis of mesial temporal lobe sclerosis is possible by demonstrating atrophy of the mesial temporal lobe structures on T1-weighed anatomical MRI images and increased signal on conventional spin-echo T2 MRI sequences (Jackson et al., 1990; Duncan et al., 1996; Woermann et al., 1998). Since hippocampal sclerosis seems to be diffuse rather than focal in most of the cases (Kim et al., 1995), an involvement of the amygdala by the pathological process underlying hippocampal sclerosis might be expected, and indeed is reported in the literature (Hudson et al., 1993; Miller et al., 1994).

In vivo identification of amygdala sclerosis by measuring the amygdala T2 relaxation time has been reported in patients with temporal lobe epilepsy (Van Paesschen et al., 1996; Kalviainen et al., 1997). Furthermore, amygdala volumetry has been validated as a reliable method (Watson et al., 1992; Cendes et al., 1993; Soininen et al., 1994; Kalviainen et al., 1997).

Rationale for this study

The aim of our study was to investigate amygdala pathology in patients suffering from temporal lobe epilepsy and additional affective aggression, specifically IED. In particular, we hypothesized that, in patients with temporal lobe epilepsy and intermittent affective aggression, amygdala sclerosis in the context of hippocampal sclerosis would be more common than in control patients. Furthermore, we wanted to test if there is an association between aggression, on the one hand, and hippocampal sclerosis, low IQ and poor social adjustment, on the other hand, in patients with temporal lobe epilepsy.

Patients and methods

Patients and patient assessment

Approval for this study had been obtained from the ethics committee of the National Hospital for Neurology and Neurosurgery. Patients with temporal lobe epilepsy were recruited from a tertiary referral centre (National Hospital for Neurology and Neurosurgery and the associated Chalfont Centre for Epilepsy). The clinical syndrome of interest was defined as complex partial seizures with a symptomatology, EEG and MRI findings compatible with temporal lobe epilepsy. Neurologists who were not involved in this study made the neurological diagnoses. On the basis of the discharge summaries, patients with temporal lobe epilepsy with and without a history of aggression were identified, contacted and seen by a psychiatrist (L.T.v.E.). Patients with extratemporal or generalized epilepsy were excluded, as were those with a history of mental handicap or psychoses. Informed consent was obtained from the patients prior to further investigations. Patients with temporal lobe epilepsy with and without a history of IED diagnosed according to DSM-IV criteria were included in the study. A neurological and psychiatric history and examination were obtained, as well as routine EEG investigations and neuropsychological investigations. Full, verbal and performance IQ were measured using the well-established revised version of the Wechsler Adult Intelligence Scale (Nachson, 1991). Patients with a full IQ below 70 were excluded from the study to avoid selection bias. Frequency and severity of the three main seizure types were documented and rated according to the National Hospital Seizure Severity Scale (O'Donoghue et al., 1996). All patients were asked to fill in the Beck's Depression Inventory and the State Trait Anxiety Inventory. These questionnaires are self-rating instruments for depression and anxiety, respectively (Thomson, 1989a, b). In order to assess aggression, carers were asked to fill in the Social Dysfunction and Aggression Scale (SDAS-21), a well-validated and recommended questionnaire (European Rating Aggression Group, 1992; Mak and De Koning, 1995).

Twenty healthy volunteers, matched for age and sex, were scanned and measured twice in order to assess the reliability of the volumetric measurements.

MRI assessment

The MRI images were obtained at the Chalfont Centre for Epilepsy on a 1.5 T GE Signa scanner (GE Medical Systems, Milwaukee, Wis., USA) using a T1-weighted inversion-recovery prepared volume acquisition [IRSPGR: TI/TR/TE/flip = 450/15/4.2/20; 124 × 1.5 mm thick contiguous coronal slices; matrix 256 × 192, 24 cm × 18 cm FOV (field of view) (TI = inversion time; TR = repetition time; TE = echo time)]. For computation of T2 values, conventional spin-echo sequence [TR/TE1/TE2/NEX 2000/30/120/1, 256 × 192 matrix, 24 × 18 cm FOV, 5 mm thick coronal slices with no gap; scan time 10 min (NEX = number of excitations)] was obtained by two interleaved acquisitions to cover the entire brain. In the latter, the slices were acquired in a tilted coronal plane perpendicular to the long axis of the hippocampi.

Volumetric measurements

Volumetric measurements were performed using the locally developed interactive software program Mreg (available on the internet: http://www.erg.ion.ucl.ac.uk/MRreg.html) (Lemieux et al., 1998; Moran et al., 1999). The images were zoomed to a magnification of ×4 for outlining of the amygdala, and intensity windowing was monitored consistently. The amygdala were outlined manually following the established protocol described by Watson (Watson et al., 1992). The intracranial volume was measured by manual delineation of the internal face of the cranium at every 10 slices, with a magnification of ×2. The volume of the delineated amygdala in each slice (the in-slice volume) was calculated by multiplying the number of voxels contained within each trace (corrected for magnification) by the voxel volume times 10. The total volume of each amygdala was the sum of all in-slice volumes. The amygdala volumes were corrected for total brain size by division by the intracranial volume (Cendes et al., 1993). The rater (L.T.v.E) was blind to the subject grouping. Intrarater reliability figures were calculated from repeated measurements of the subset of 20 normal controls.

Amygdala atrophy was defined as a volume <3 SD (standard deviations) below the average amygdala volume of the control group. This criterion was preferred to the less restrictive threshold of 2 SD, more commonly applied to hippocampal measurements (Cook et al., 1992; Van Paesschen et al., 1995) in order to account for the larger variability in the amygdala measurements due to greater difficulty of the measurements.

Amygdala T2 (AT2) mapping

Pixel-by-pixel T2 maps were calculated from the images obtained with the spin-echo sequence, using the expression T2 = (TE2 – TE1)/[ln (S1/S2)], where S1 and S2 are the signal intensities in the early and late echo images, with echo times TE1 and TE2, respectively. This dual-echo technique has been validated including repeated acquisitions of data and measures of intra- and inter-rater and test–retest reliability in hippocampi of controls and patients (Duncan et al., 1996; Woermann et al., 1998). AT2 values were measured by the same observer (F.G.W.) using DispImage image analysis software (Plummer, 1992) by placing the largest possible elliptic region of interest within the amygdala while avoiding anatomical boundaries. The coronal slice for AT2 measurements was defined as the one anterior to the slice in which the hippocampal head was seen last underlying the amygdala. Intra- and inter-rater variability of AT2 measures using this method were assessed by calculation of the limit of agreement and coefficient of repeatability (Bland and Altman, 1986). AT2 signals were defined as pathological if they were >2 SD above the mean of the control population.

Data analysis


The intrarater reliability was assessed by three different methods to enable comparison with published figures and to provide conservative figures for further comparisons: (i) the ratio of measured standard deviations to average amygdala volumes; (ii) the coefficient of repeatability, Cr (Bland and Altman, 1986); and (iii) an intraclass correlation coefficient (Streiner and Norman, 1995).

Group comparisons

In order to test our main hypotheses of increased amygdala pathology in aggressive patients, we compared the amygdala volumes and amygdala T2 relaxation times of patients with temporal lobe epilepsy with and without IED using two-tailed standard t-tests. We then compared categorical data such as laterality and nature of EEG abnormality and MRI pathology between the two groups using contingency tables. In order to examine the influence of the different categories of the tables, we used closed test procedures (Toothaker, 1992; Horn and Vollandt, 1995). Categories of the variables that contribute significantly to the overall significance of the table are flagged in the case of P < 0.05 and double flagged in the case of P < 0.01 (Tables 2 and 3). We then compared the IQ and psychometric results between the two patient groups using standard two-tailed t-tests. We corrected the significance of these findings using the Bonferroni method (Norman and Steiner, 1993), and again flagged the findings in the case of P < 0.05 and double flagged in the case of P < 0.01 (Table 4). All data were analysed using SPSS for Windows (Release 7.5.1).


Study group structure

Forty-three patients with epilepsy and a history of aggressive behaviour were identified; 25 fulfilled inclusion criteria and 18 were excluded for different reasons (post-surgical n = 2; history of mental retardation or IQ < 70 n = 4; psychosis n = 2; phenomenology of aggression, i.e. history of antisocial personality disorder, n = 2; classification of epilepsy n = 8). A quantitative MRI scan was obtained from 24 of these patients. One patient could not be scanned as he was using a vagal nerve stimulator.

Thirty-nine patients with temporal lobe epilepsy without a known history of aggressive behaviour were identified and contacted. Twenty-five of these patients were included in the study and 14 were excluded (11 because of a history of temper tantrums and four because of a history of ictal aggression on careful behavioural assessment). Quantitative MRI scans were obtained from 24 patients. One patient did not comply with the MRI procedure due to claustrophobia.

The demographic data of both groups are summarized in Table 1. The two patient groups were matched for age, sex, demographic background, duration of epilepsy and seizure severity.

View this table:
Table 1

Descriptive statistics: comparison of demographical and historical data of patients with temporal lobe epilepsy with (AGG) and without (non-AGG) IED

Age [range] (years)30.1 [18–49]33.8 [19–56]
Sex: F/M 8/1710/15
Work: total out of 25 unemployed2115
Living: total out of 25 living independently1410
Income: total out of 25 on social support1816
Social: total out of 25 living in stable relationship 8 8
Therapy: monotherapy– polytherapy 3–22 3–22
Mean duration of temporal lobe epilepsy [range] (years)22.4 [5–45]24.5 [7–46]
Mean seizure frequency [range] (estimated frequency per month)13.4 [0.5–60]21 [1.5–190]
Birth complications 9 7
Febrile convulsions 5 9
Status epilepticus 0 2
Left handed 7 1
History of encephalitis 5 0

There was no significant group difference regarding the history of birth complications, febrile convulsions or status epilepticus. However, the incidence of encephalitic brain disease (Fisher's exact test: P = 0.05) and left handedness (χ2 test: P < 0.05) was significantly increased in aggressive patients (see Table 1).

Reference and reliability data

Amygdala volumes and intracranial volumes were measured in 20 healthy controls using the method described above. The mean right amygdala volume (1906.7 mm3; SD = 172.4 mm3; SE = 38.8) was non-significantly smaller than the mean left amygdala volume (1913.0 mm3; SD = 162 mm3; SE = 36.2). The coefficient of variation expressed as a percentage of the standard deviation of the mean volume was 8.7% which compares well with figures published in the literature (Soininen et al., 1994; Kalviainen et al., 1997). The coefficient of repeatability was Cr = 295.7 mm3, i.e. 15% of the overall mean. The intraclass correlation coefficient calculated as described above was 0.7, which again corresponds well to what has been reported in the literature (Öngür et al., 1998).

Intrarater repeatability of AT2 in the control group showed a mean difference of 1.2 ms (1.4% of the mean AT2), a limit of agreement of 4 ms and a coefficient of repeatability of Cr = 4.7%.

EEG abnormalities

As shown in Table 2, there was less right-sided focal EEG abnormality and more bilateral EEG abnormality in the aggressive group compared with non-aggressive patients with temporal lobe epilepsy.

View this table:
Table 2

Laterality of EEG abnormalities

AGGNon-AGGGroup comparison
Overall significance: P = 0.03; closed test procedure: *P < 0.05. R = right; L = left; AGG = aggressive group; non-AGG = non-aggressive group.
No abnormality 2 0
R focal abnormality 2 9*
L focal abnormality1010
Bilateral focal abnormality11 5*

MRI pathology

Radiological diagnosis

Table 3 summarizes the radiological diagnoses for the two patient groups. Left- as well as right-sided hippocampal sclerosis was significantly less common in patients with temporal lobe epilepsy and IED. Other left temporal pathology, including three patients with amygdala pathology (amygdala sclerosis, amygdala glioma and amygdala dysontogenetic neuroepithelial tumour), two patients with multiple small temporal infarctions and two patients with diffuse left temporal atrophy of unknown origin, was significantly more common in patients with temporal lobe epilepsy plus IED.

View this table:
Table 3

Radiological MRI diagnoses

AGGNon-AGGGroup comparison
Overall significance: P = 0.002; closed test procedure: *P < 0.05, **P < 0.01. Other left temporal pathology: amygdala sclerosis = 1, amygdala glioma = 1, amygdala dysontogenetic neuroepithelial tumour = 1; multiple small temporal infarctions = 2; generalized left temporal atrophy = 2]; HS = hippocampal sclerosis; R = right; L = left; AGG = aggressive group; non-AGG = non-aggressive group.
No pathology10 6
R HS 1 8**
L HS 410**
Bilateral HS 3 1
Other left temporal pathology 7 0**

AT2 measurements

There was no evidence of an increased AT2 relaxation time in the aggressive group [right side: AGG (aggressive group) = 85.8 ms, SD = 4.74 m; non-AGG = 87.83 ms, SD = 6.83 ms; left side: AGG = 85.88 ms, SD = 5.23 ms; non-AGG = 87.03 ms, SD = 4.31 ms). Using the definition of amygdala T2 pathology as a T2 time > 2 SD above the mean of the control group, there was no significant group difference in amygdala pathology.

Amygdala volumes

A group comparison did not reveal a significant overall difference in amygdala volumes (right side: AGG = 1893 mm3, SD = 435 mm3; non-AGG = 1909 mm3, SD = 231 mm3; left side: AGG = 1840 mm3, SD = 398 mm3; non-AGG = 1868 mm3, SD = 290 mm3). However, in the aggressive patients, a subgroup of five patients (20%) showed amygdala atrophy (χ2: P = 0.04). Two patients had left-sided amygdala atrophy, two had bilateral atrophy and the only patient who exhibited right-sided amygdala atrophy was left handed.

Comparing patients with and without amygdala atrophy, there was a highly significant group difference in amygdala volumes for both sides (right side: atrophy = 1336 mm3, SD = 183 mm3; no atrophy = 1982 mm3, SD = 281 mm3; P < 0.000; left side: atrophy = 1285 mm3, SD = 431 mm3; no atrophy = 1935 mm3, SD = 244 mm3; P < 0.000). Other than this difference, an increased incidence of encephalitis (χ2 test: P < 0.05; Fisher's exact test: P = 0.1) was the only clinical feature that distinguished patients with amygdala atrophy from those with normal amygdala volumes.

Neuropsychological profile

There was a highly significant group difference in IQ figures, with the verbal IQ, the performance IQ and hence the full IQ all being lower in the aggressive group (Table 4). The verbal IQ differed more than the performance IQ.

View this table:
Table 4

Neuropsychological and psychometric parameters

VariableAGGNon-AGGGroup comparison
*P < 0.05, **P < 0.01 after Bonferroni correction; FIQ = full IQ; VIQ = verbal IQ; PIQ = performance IQ; BDI = Beck's Depression Inventory; STAI = State Trait Anxiety Inventory; SDAS = Social Dysfunction and Aggression Scores.
Mean FIQ (SD)80.6 (8.5)93.0 (13.9)**
Mean VIQ (SD)81.0 (8.6)93.8 (14.1)**
Mean PIQ (SD)83.3 (11.8)94.7 (15.1)*
Mean BDI (SD)8.8 (4.8)4.2 (5.8)*
Mean State STAI (SD)44.1 (14.9)32.1 (10.9)*
Mean Trait STAI (SD)50.1 (10.1)34.7 (11.6)**
Mean SDAS-9 (SD)14.9 (7.5)0.4 (0.7)**
Mean SDAS-21 (SD)30.9 (12)3.1 (4.5)**

Psychometric profile

As expected, there was a highly significant group difference in the Social Dysfunction and Aggression Scores (SDAS 9, SDAS 21) since this was the criterion for group definition (Table 4). There was a significant group difference in Beck's Depression Inventory and the State Trait Anxiety Inventory scores, with the aggressive group rating much higher in depression (P < 0.05), State (P < 0.05) and Trait anxiety (P < 0.01).


This is the first study investigating a possible link between amygdala pathology and aggressive behaviour in temporal lobe epilepsy using quantitative MRI. Before embarking on the interpretation of our findings, we want to address a number of methodological issues.

Methodological issues

Reliability of volumetric measurements

The MRI quantification methodology used has been shown to be sensitive for the detection of mesial temporal sclerosis including amygdala sclerosis (Van Paesschen et al., 1996), and our reliability figures are similar to those published (Soininen et al., 1994; Kalviainen et al., 1997; Öngür et al., 1998).

Patient selection

All patients suffered from chronic medically intractable temporal lobe epilepsy and were identified at a tertiary referral centre. Subjects with a history of major psychiatric disorder were excluded. The two patient groups were homogeneous in terms of demographical background and clinical features relating to the epilepsy.

Behavioural assessment

We included only patients with a clearly defined psychopathological syndrome according to DSM-IV criteria, i.e. IED. Patients with other forms of aggressive behaviour not fulfilling these criteria, such as patients with antisocial personality disorder, were excluded. Thus our study group is homogenous in terms of psychopathology.

Neuropsychological assessment

We have excluded patients with a history of mental handicap. Following neuropsychological assessment, we also excluded patients with a full IQ outside the normal range, i.e. < 70. Thus we assert that there was no neuropsychological selection bias and that the group differences in IQ relate to the aggressive syndrome of interest.

Assessment of brain pathology

The main inclusion criteria were the clinical diagnoses of temporal lobe epilepsy and IED without regard to the underlying brain pathology. Following patient inclusion, we obtained the MRI and neurophysiological data. Thus, no selection bias should exist and differences in brain pathology should be associated with the difference in psychopathology.

Intermittent explosive disorder in temporal lobe epilepsy

A subgroup of 20% of patients with IED suffered from very severe amygdala atrophy. In this group, the mean amygdala volumes of both sides were significantly smaller than those of patients without amygdala atrophy, which is not surprising since amygdala atrophy was the group-defining variable for this comparison. However, we did not find increased T2 relaxation times as an indicator of amygdala sclerosis. None of these patients was diagnosed as suffering from amygdala sclerosis on visual assessment. A history of encephalitis was more common in these patients, and one might speculate, therefore, that an encephalitic brain disease led to this severe amygdala atrophy. However, we cannot answer the question of whether or not encephalitis renders patients more vulnerable to the development of amygdala atrophy.

Seven of the aggressive patients (28%) suffered from different pathologies affecting the left temporal lobe, including the left amygdala, compared with none in the non-aggressive group. Only one of these seven patients had amygdala atrophy. Thus in a total number of 11 of 25 patients (44%) with temporal lobe epilepsy plus IED, there was evidence of amygdala pathology. Even though the nature of this pathology is diverse, it might well be the reason for unprovoked states of arousal and anger in these patients.

The other clear difference in brain pathology between the two groups is the different prevalence of clear-cut unilateral hippocampal sclerosis. This is the most common cause of temporal lobe epilepsy, and 18 of 25 patients with temporal lobe epilepsy alone but only five of 25 patients with temporal lobe epilepsy plus IED had this condition. Thus we could not confirm the increased incidence of hippocampal sclerosis in patients with temporal lobe epilepsy and aggression reported by Falconer (Falconer, 1973).

Bilateral focal EEG abnormalities were significantly more common in patients with temporal lobe epilepsy and IED. This finding might reflect the less focal and homogeneous brain pathology in this patient group. Right-sided focal EEG abnormalities were significantly less common in the aggressive patients and may indicate a more important role for left structural brain pathology in the subsequent development of IED. IED was strongly associated with a decreased overall IQ and in particular a low verbal IQ, reflecting dominant hemisphere functions. This finding corresponds well not only with previous studies in epilepsy but also with other studies of impulsive aggression (Mungas, 1988; Foster et al., 1993; Barratt et al., 1997; Nelson et al., 1998).

There is controversy regarding the importance of hemispheric specialization for aggressive behaviour (Bear, 1983; Nachson, 1991), with the majority of studies pointing to a more important role for the left hemisphere (Saver et al., 1996). Our MRI and neuropsychological findings (i.e. particularly low verbal IQ) support this assumption. Furthermore, the increased prevalence of left handedness in our aggressive patients may indicate early brain pathology affecting the left hemisphere (i.e. lateralization of dominance to the right hemisphere).

There was a strong link between aggression and high levels of depression and anxiety, confirming reports of such an association in the non-psychiatrically ill population (Bjork et al., 1997). It seems plausible that high levels of anxiety result in states of hyperarousal that might be facilitated by amygdala pathology, as suggested by other authors (Cendes et al., 1994). Regarding the relationship between depression and aggression, there are only a few non-conclusive reports (Braconnier and Jeanneau, 1997). Our findings point to a clear association between depression and IED in temporal lobe epilepsy.

It is interesting to compare our data with those of Herzberg and Fenwick who did not find any specific pattern of brain pathology in a retrospective study of 31 patients with temporal lobe epilepsy with and without aggressive outbursts (Herzberg and Fenwick, 1988). However, they had to rely on CT scans for the assessment of brain pathology. In agreement with our study, they found a lower mean IQ and poorer occupational records in their aggressive patients. They also found a higher prevalence of psychiatric co-morbidity in these patients, a finding on which we cannot comment since we excluded patients with psychiatric disorders other than IED.

While from our study the amygdala emerge as a critical structure for the mediation of aggressive behaviour, the literature suggests that there are clearly other critical structures involved, such as the periaqueductal grey, the hypothalamus or the frontal lobes. In this context, it would be of particular interest to quantify frontal lobe pathology, since from a phenomenological point of view IED includes symptoms of hyperarousal and dyscontrol at the same time, and frontal lobe pathology is known to be associated with dyscontrol symptoms.

Clinical relevance of the findings

First, it is interesting to note, and reassuring for patients with this condition, that hippocampal sclerosis is not associated with the development of IED. Second, while the causal relationship between low IQ, high anxiety and high depression in those patients with temporal lobe epilepsy and IED remains unclear, our findings might indicate a possible therapeutic intervention. For example, the treatment of depression and anxiety might help these patients and their carers to cope more effectively with everyday life and to avoid hyperarousal states leading to aggressive behaviour. Finally, early treatment of possibly reactive psychiatric symptoms such as anxiety or depression in patients with temporal lobe epilepsy may have prophylactic properties for the prevention of aggressive behavioural disorders.


In summary, our results indicate that hippocampal sclerosis and the pathogenic process leading to it do not predispose patients with temporal lobe epilepsy to the development of interictal IED. IED is associated with more distributed and predominantly bilateral or left-sided brain pathology. There was no evidence of amygdala sclerosis in the aggressive group in terms of increased mean T2 relaxation times. There was a significant subgroup of 20% of patients with temporal lobe epilepsy and aggression, who displayed very severe amygdala atrophy, with volumes < 3 SD below the mean. A history of encephalitis was more common in these patients and might have been the cause for this severe atrophy. Another subgroup of 28% of aggressive patients displayed different left temporal lobe lesions affecting the amygdala or periamygdaloid areas. Dominant hemisphere functions seem to play a more important role in the genesis of IED, with right-sided EEG and MRI abnormalities being less common in patients with temporal lobe epilepsy and aggression. IED, in patients with temporal lobe epilepsy, is closely linked with a decreased intelligence, in particular verbal intelligence, and increased anxiety and depression. Treatment of anxiety and depression might help patients and carers to cope more effectively with everyday life and, by avoiding hyperarousal states, might help to improve the disturbed behaviour.


We wish to thank Dr S. Free, Dr S. Brocklehurst and Professor J. Schulte-Mönting for their methodological and statistical advice, and Professor J. S. Duncan, Dr D. Fish, Professor W. S. Sanders, Professor S. S. Shorvon, Dr S. Krishnamoorthy, Dr C. Selai and R. Linney for their support of this study. This project was supported by the Wissenschaftliche Gesellschaft, Albert-Ludwigs-University of Freiburg, Germany and the Raymond Way Fund, Institute of Neurology, UCL, London, UK.


View Abstract