The neurobiology of violence in patients with schizophrenia is becoming an area of increasing academic interest. In their detailed review of neurobiological correlates of violent behaviour in patients with schizophrenia, Naudts and Hodgins [1] point out that evidence has accumulated showing that, compared with the general population, patients with, or who will develop, schizophrenia are at increased risk for violent offending and at even higher risk of committing homicide, with such evidence being derived from investigations of birth and population cohorts comparing the criminality of those with and without schizophrenia, from follow-up studies of schizophrenia patients in the community, from diagnostic studies of representative samples of incarcerated offenders and from investigations of complete cohorts of homicide offenders.
Only a handful of structural neuroimaging studies have been carried out so far to examine whether or not there is a systematic difference in brain grey matter between patients with schizophrenia who have carried out acts of serious violence and patients with schizophrenia who have not done so.
The first published study was that of Chesterman et al. [2], in which 10 male inpatients of a special (high-security) hospital in the south of England underwent cerebral magnetic resonance imaging (MRI). All had committed at least one act of very serious violence, or similarly dangerous acts (e.g. arson). This study reported that the majority of these patients showed reduction in mesial temporal structures. However, not all the patients were diagnosed as suffering from schizophrenia; only six had such a diagnosis, while the remaining four were diagnosed as suffering from a primary personality disorder. Furthermore, there was no direct comparison group.
In the American computed tomography (CT) study by Convit et al. [3], nine male inpatients with schizophrenia with had a history of repetitive violence were compared with nine male inpatients with schizophrenia who had no history of violence. No significant differences were found in cortical atrophy between the two groups, but the Sylvian fissure (lateral sulcus) was subjectively rated as being larger bilaterally in the violent group. No information was available regarding current antipsychotic medication.
In the MRI study of Wong et al. [4], 31 inpatients from a British maximum security hospital were compared with eight normal controls. Of these 31 inpatients, 17 had a history of repetitive violence; they showed a reduction in the volume of the amygdala compared with the normal controls. Similarly, 14 had a history of having committed one violent offence and also showed a reduction in the volume of the amygdala compared with the eight normal controls. Unfortunately the inpatient sample chosen did not have a clear-cut diagnosis of schizophrenia; they were recorded as suffering from either schizophrenia or schizoaffective disorder. Furthermore, it is unclear how many of these inpatients were being treated with antipsychotic medication at the time of the neuroimaging study; presumably the comparison group of eight healthy controls were antipsychotic-naïve.
Barkataki et al. [5] also carried out an MRI study of 13 male inpatients from a British maximum security hospital who had been detained in secure conditions because of their violence (homicide, attempted murder, wounding, robbery, or 'other type of serious violence'). They were compared with a group of 15 male patients with schizophrenia who had no history of serious violence. There was no significant difference in the mean chlorpromazine-equivalent antipsychotic dosage received by each group. Stereological volumetric assessment was conducted using the MEASURE program and Cavalieri method [6], in which a three-dimensional grid of voxel points was overlaid onto a reconstructed three-dimensional MR image that had been reoriented parallel to the anterior-posterior commissure and inter-hemispheric fissure. Using point counting, a rater then manually assessed the volumes of the whole brain, cerebellum, temporal lobe, lateral ventricles, caudate nucleus, putamen, thalamus, hippocampus and amygdala, blind to group. Using this technique, the whole brain was found to have a lower volume and the putamen and amygdala were found to have higher volumes in the violent group; all other comparisons between the two groups were not statistically significant. A further study was carried out on 12 of the 13 inpatients with a history of violence compared with the 15 patient without such a history, to study the cerebral cortex using a cortical pattern-matching method following the manual delineation of 31 sulcal landmarks in each hemisphere of the segmented brain (in which the cerebellum as well as non-brain tissues were removed before analysis) [7]. The violent patients were reported as showing thinning in the right sensorimotor cortex (right M1 and S1) compared with the non-violent patients. A difficulty with these two studies relates to the fact that, in both groups, there was evidence of comorbidity with substance abuse. In particular, in the violent schizophrenia group, two patients had a history of alcohol dependence, one of solvent misuse, two of cannabis dependence, one of alcohol as well as polysubstance misuse (cannabis, ecstasy, LSD and amphetamines) and one of alcohol dependence and polysubstance misuse.
Hoptman et al. [8] manually traced the orbitofrontal cortex on anatomical images from MRI scans carried out at two different American centres as part of a double-blind treatment study comparing the response to four different antipsychotic drugs in chronic antipsychotic treatment-resistant patients. Although imaging parameters and pixel dimensions varied by site, volumes were reported in cubic millimetres, which allowed direct comparisons of data from a total of 49 useable scans across sites. Larger left orbitofrontal cortex grey matter volumes were reported to be associated with greater levels of aggression, recorded using the Overt Aggression Scale, which provides subscales for verbal aggression, physical aggression against objects, physical aggression against self, physical aggression against other people and intervention [9]. Unfortunately, not all the patients had a diagnosis of schizophrenia; some were diagnosed as suffering from schizoaffective disorder, but the results of the scan data from patients with both diagnoses were combined in the publication of the results from this study. Furthermore, over half the sample had a comorbid alcohol use diagnosis (dependence or abuse) and 59% had a comorbid substance use diagnosis.
It can be seen that there have been disadvantages in all these previous studies. These have included, variously: a lack of an appropriate control group; a lack of information about antipsychotic medication, the effect of which on grey matter cannot be ruled out; a lack of a clear-cut single diagnosis of schizophrenia unaccompanied by comorbid psychoactive substance misuse (the effect of which on grey matter cannot be ruled out) or the inclusion of schizoaffective disorder in the group(s) studied; the exclusion of the cerebellum in the analysis; and the use of manual tracing methodologies, which might introduce bias and are likely to have less than a perfect inter-rater reliability. For the present study, therefore, we chose to examine patients with a clear-cut diagnosis of schizophrenia, who did not suffer from any comorbid psychiatric disorder (such as psychoactive substance misuse), and for whom information about antipsychotic medication was available, dichotomized into two groups, one of which had a history of serious violence and the other of which did not. In order to study the grey matter of the brain, including the cerebellum, from MRI scans, without introducing manual tracing methods, we chose to use the unbiased approach afforded by voxel-based morphometry, which requires no a priori information about the location of possible differences between groups and which is not operator-dependent. The technique involves spatially normalizing all the images to the same stereotactic space (by registering each of the images to the same template image, by minimizing the residual sum of squared differences between them), segmenting the grey matter from the normalized images, correcting for volume changes arising from spatial normalization and, finally, carrying out a statistical analysis to localize differences between groups; the output from the method is a statistical parametric map which shows regions where grey matter concentration differs significantly between groups [10, 11].
We report the first voxel-based morphometry study of grey matter changes in the whole brain in schizophrenia associated with a history of seriously and violently offending.