The Edinburgh High Risk Study: current status and future prospects

Schizophrenia is a relatively common disorder with a lifetime risk of eight per thousand (1) and for the last 30 years extensive investigations have been conducted to try to understand the underlying mechanisms which give rise to the disease. It is known that genetic factors are important and it has widely been hypothesised for the last 15 years that schizophrenia is a disorder of neurodevelopment which is largely compensated until early adult life (2). Direct evidence for this is not yet available and could only be obtained if it were possible to examine people who are at risk to develop schizophrenia, at some stage before they become ill, comparing them with normal controls, and follow them through the period of risk of schizophrenia until they do or do not develop the psychosis.

Schizophrenia is not sufficiently common a disorder for such a design to be practicable in the general population. Prospective studies concerning people of enhanced risk (high risk studies) have been conducted for many years, for example the Copenhagen high risk project (3), the Israeli high risk study (4) and the New York high risk project (5). These studies have identified high risk individuals in infancy as the children of schizophrenic mothers. Such a design has many practical difficulties, largely because of the 20 year gap between the identification of the subjects and their entry to their period of maximum risk of the onset of schizophrenia.

The Edinburgh High Risk Study seeks to address these issues by investigating individuals who are at enhanced risk because they have two or more affected relatives, but who are identified in their later teens, just before they enter the period of maximum risk for the condition, and the question of whether they will develop schizophrenia is likely to be resolved one way or another within 10 years. This means that the whole study can be conducted within a 10 year framework, so that attrition of the sample is likely to be less than has been the case with the high risk studies concerning people identified as infants. A further advantage of the relatively short timeframe of the study is that the instruments and methods used to investigate the population from the point of their ascertainment will not be seriously out of date by the time the study is completed. The investigation is wideranging, involving psychopathological, psychological, social, imaging and other assessments. The study began in 1994 and will run until 2004, so that it is not yet complete and the final comparison of the baseline variables between control subjects, those high risk subjects who do develop schizophrenia and those who do not, cannot yet be carried out. Significant differences between the subject groups have, however, already been established for many of the comparisons made and detailed findings of the individual areas of work have been presented. This paper reviews the method of the study and the main findings obtained so far.

METHODS

Derivation of the sample

High risk individuals were defined as young people aged between 16-25, who did not have a history of serious psychiatric problems and had never been considered as psychotic, who had at least two first or second degree relatives affected with schizophrenia. This was done by examining casenotes of all patients with schizophrenia known to individual hospitals where there appeared to be two related cases. Consent was sought from one of the affected subjects to speak to a healthy relative, from whom a full family history was obtained, particular emphasis being given to the possibility of their being family members, aged 16-25, who were first or second degree relatives of the affected subjects. This was a labour-intensive process involving the scrutiny of over 2000 sets of casenotes and home visits to more than 500 potential subjects. When the investigation was planned, we were able to calculate the likelihood of individuals, from some specific families known to us, of developing schizophrenia by the age of 30, which was between 10 and 15%. We, therefore, determined that we should seek to obtain 200 high risk subjects with a view to 20-30 of them developing schizophrenia. There are two control groups, each planned to be of about 30 subjects, to compare with those who will develop schizophrenia, one of age-matched individuals with no family history of psychotic disorder and one of age-matched cases of first episode schizophrenia who were not known to have a family history.

Two hundred and twenty nine suitable high risk individuals who were prepared to consider inclusion in the study were identified and, in the first five years of the study, 162 of these provided useful data. In order to achieve these numbers we included mental health services over much of Scotland, so that, in addition to having substantial numbers of subjects from the city of Edinburgh, we have included people from families living in the rural areas of Argyll, Clyde, Borders, Forth Valley, Lothian, Highlands & Islands and the towns and cities of Inverness, Dumfries, Perth and Greenock. These are areas of stable population where traditional patterns of family life largely persist, providing the extended family networks necessary for the study. As subjects are from such diverse environments, we have recruited the well controls from the social networks of the subjects themselves. The first episode cases were recruited from local hospitals and were balanced group-wise for age (and sex as far as possible) with the high risk subjects. The demographics and baseline social characteristics of the three groups studied are shown in Table 1.

Table 1.

Baseline characteristics of the study groups

	High risk	Well controls	First episode schizophrenics
No. of cases	76M; 86F	17M; 19F	25M; 12F
Age (years, mean ± SD)	21.19 ± 2.97	21.17 ± 2.37	21.63 ± 3.69
Social class (%, based on paternal occupation)
1 or 2	18.5	30.6	27.0
3 or 4	53.7	47.2	35.0
5 or 6	24.7	16.7	8.1
Unclassifiable	3.1	5.6	29.7

Plan of study and assessments used

The study has been conducted in two phases. The first took place from 1994 to 1999 and the second will run from 1999 until 2004. The plan of the first phase was to assess all of the high risk subjects and both groups of controls at the point of ascertainment in terms of psychopathology, neuropsychology and brain structure as determined by structural magnetic resonance imaging (MRI). The first episode cases were assessed only on ascertainment while in the high risk subjects and in the well controls the assessments were repeated every 18 months. In addition to these measures, we also assessed childhood behavioural traits, schizotypal features, minor physical anomalies, dermatoglyphics, life events and substance use. Our main psychopathological instrument was the Present State Examination, ninth edition (PSE-9) (6) and, for simplicity, psychopathology was graded in terms of the following scale:

• 4 -definite schizophrenia
• 3 -any fully-rated psychotic feature from the specific categories of delusions, hallucinations and passivity experiences
• 2 -partially-rated symptoms as described in 3 and fullyrated other perceptual disorders
• 1 -no psychotic or possibly psychotic features but definitely recorded features in other categories such as depression and anxiety
• 0 -none of the above.

An extensive battery of neuropsychological tests was conducted, which has been described in detail elsewhere, but essentially consisted of tests of general intelligence, attention, motor speed, executive function, verbal learning and memory. Brain structure was assessed using structural MRI. Scans were initially assessed volumetrically, but have subsequently been re-assessed using voxel-based methods. Elements of childhood personality and behaviour, life events, minor physical anomalies, and other variables were assessed using standardised methods. We were also able to assess the degree of genetic liability that the high risk subjects had. This was done by two methods. The first was a categorical one, considering the numbers of relatives of first or second degree known to be affected, so that the categories were:

• 0 -no relatives affected, i.e. control subjects
• 1 -second degree relatives only
• 2 -one first degree and one or more second degree relatives
• 3 -two or more first degree relatives.

This method does not take account of the entire numbers of relatives that the subjects had and so we also developed a continuous method of genetic liability (described in Lawrie et al. [7]).

In the second phase (1999-2004) of the study, psychopathological assessments have continued and increasing numbers of subjects are gradually developing symptoms and becoming floridly unwell. We have been able to keep contact with a high proportion of the sample and serial functional MRI scans are being successfully conducted, but full results are not yet available.

RESULTS

It is evident from Table 1 that the three groups of subjects are well matched, apart from the fact that there were fewer females in the first episode group. This would be expected in a sample of this age, as females have a slightly later age of onset than males (8). The unclassifiable social class of some of the schizophrenic control group is due to the fact that we were less well informed about the families of those subjects than of the other two groups. The difference in social class is not significant, but the finding that high risk subjects appear to be of somewhat lower occupational class is likely to be due to the fact that in a number of cases the fathers had schizophrenia and had suffered occupational decline.

It is evident from Table 2 that symptoms of all kinds, including psychotic symptoms, have occurred in both the control and the high risk group, but that they have been more frequent in the high risk sample. Schizophrenia itself has only been diagnosed in the high risk subjects (9). Eighteen of the high risk subjects have now developed schizophrenia and none of the controls has, although three control subjects did show a Category 3 PSE rating (i.e., they had isolated psychotic symptoms). Such symptomatology has not been sustained in the control subjects, who are all free of psychotic symptoms now. We are successfully maintaining contact with most subjects, both high risk and control, even though the study has now been in progress for more than seven years (Table 3).

Table 2.

Highest Present State Examination (PSE) rating ever obtained in high risk subjects and well controls

Highest score ever obtained	High risk (valid %)	Well controls (valid %)
on PSE	(n=162)	(n=39)*
0	40 (25.8%)	26 (64.8%)
1	36 (23.2%)	8 (21.1%)
2	51 (32.9%)	1 (2.6%)
3	10 (6.5%)	3 (7.9%)
4	18 (11.6%)	0

* Three extra control cases were recruited in the second 5-year phase of study, as it initially appeared that we might not be able to keep contact with control subjects. Contact was achieved in the end (see Table 3)

Table 3.

Current status of all subjects recruited for the study

Current status	High risk	Well controls
	(n=162)	(n=39)
PSE data complete	101	26
General practitioner says well	16	2
Withdrawn	9	1
Relative/individual reports well	(8)
Status unknown	(1)
Current and future appointments	15	3
Travelling/unavailable	1	-
Lost contact	20 *	7 *

* These numbers are still being reduced with active tracing methods

We plan to see all subjects for a final assessment in 2003. In terms of the assessments carried out so far, Table 4 shows that, in addition to the differences in psychopathological findings described above, relatively poor neuropsychological performance is widespread in the high risk subjects. In some tests, this impaired performance is related to the degree of genetic liability, those with greatest genetic liability showing worse performance. This impairment of neuropsychological performance occurs in more subjects than are ever expected to develop schizophrenia (10,11). Those who do go on to develop psychotic symptoms show impairments within the range of the other high risk subjects at the beginning, but they deteriorate further as psychosis approaches (12). In terms of brain structure, high risk subjects have reduced amygdalo-hippocampal and thalamic volumes in comparison with controls from the outset of the study. Some of the structural changes relate to the degree of genetic liability, such that there is a greater deviation from control findings in those whose genetic liability is greater (7,13). In those subjects who go on to develop psychotic symptoms there are reductions in temporal lobe volumes (14). Thus, as far as both neuropsychology and brain structure are concerned, deviations from the normal are seen in more subjects than are predicted to develop schizophrenia and, to some degree, these deviations relate to genetic liability. As psychosis approaches, further deteriorations occur.

Table 4.

Principal findings so far

Psychopathology	All areas of psychopathology more frequent in high risk subjects than controls; schizophrenia only developing in high risk subjects
Neuropsychology	High risk subjects performed less well than controls in all tests of intellectual function and on aspects of executive function and memory, some of the latter being related to genetic liability. Further decline in function in those with psychotic symptoms
Brain structure as determined by magnetic resonance imaging	High risk subjects have reduced amygdalo-hippocampal and thalamic volumes compared to controls and larger amygdalo-hippo-campi than schizophrenics
	Prefrontal and thalamic volumes relate to degree of genetic liability. High risk subjects with psychotic symptoms show reduction in temporal lobe volumes

From Table 5 it is evident that measures of personality and behaviour, whether assessed with the patient at interview, on the basis of a maternal account, or on the basis of the subject's self-rating, are predictive of the development of psychotic symptoms and of schizophrenia (15-17). High risk subjects show more frequent abnormalities of neurodevelopment, including hypertelorism, than the control group, but these measures have no association with genetic liability (18,19). Genetic liability is clearly important, but within this sample, which has been selected on the basis of exceptionally high genetic liability, other non-genetic factors, namely illicit drug use and life events, are relevant to the development of psychotic symptoms (20). Genes, therefore, are not everything even in a sample such as this.

Table 5.

Additional findings

Childhood behavioural traits (maternal rating)	Various behaviours, particularly withdrawn, aggressive behaviour, aged 13-16, are significant predictors of the later development of psychotic symptoms and schizophrenic illness
Schizotypal features (assessed by interviews with psychiatrist)	A combination of schizotypal features, including social withdrawal, psychotic symptoms, socio-emotional dysfunction and odd behaviour, is predictive of schizophrenia. On an individual basis, these features are not predictive
Self-rated schizotypal cognitions	Total scores highly associated with the presence of psychotic symptoms and high scorers at increased risk of schizophrenia
Neurodevelopmental indices	Minor physical anomalies more frequent in high risk subjects than controls.
	No relationship with psychotic symptoms or genetic liability
Hypertelorism	Interlens and interorbital distance greater in first episode schizophrenics and high risk subjects than controls, but not associated with psychotic symptoms or genetic liability
Illicit drug use and life events	Cannabis and other illicit drugs associated with psychotic symptoms in high risk subjects and controls. Major life events associated with psychotic symptoms

DISCUSSION

This is a brief summary of the findings of a long and complex study. At present our interpretation is that what is inherited in these individuals of enhanced genetic liability is a state of vulnerability to schizophrenia which does not necessary translate into psychotic illness. The results are not yet complete, as the study has not reached its end point, but it appears likely that it will be possible to accurately predict, from a sample such as this, those individuals who are and are not at particularly enhanced risk of developing schizophrenia. It is important to note that some individuals of high genetic liability are entirely well (21). It is also important to note that, at least to some extent, modifiable environmental variables such as illicit drug use and major life events are relevant to the development of illness. From the imaging results described here, and indeed from functional imaging results which are only now becoming available, it is clear that this study is going to reveal important information about the pathophysiology of schizophrenia. In addition, it is demonstrating that before psychosis develops, changes in neuropsychological performance and indeed in psychopathology occur. Some of these changes are potentially modifiable, and it may be that the study will lead us to new effective interventions. Certainly, it is suggesting new possibilities for management.