Study groups

The Edinburgh High-Risk Study is a longitudinal study of subjects at high risk of developing schizophrenia, as they have at least two first- and/or second-degree relatives with schizophrenia. They were initially recruited aged 16-24 years so that they would pass through the age at greatest risk in the subsequent 5-10 years, as previous high-risk studies of subjects recruited as children have had difficulties sustaining contact until a sufficiently high number develop psychotic symptoms. Details of the recruitment process have been described in other papers (Reference Hodges, Byrne and GrantHodges et al, 1999; Reference Johnstone, Abukmeil and ByrneJohnstone et al, 2000). Briefly, individuals with schizophrenia, with a family history of schizophrenia and with adolescent relatives, were identified from psychiatric hospital case records in many areas of Scotland. Case-note diagnoses of schizophrenia were verified with the Operational Criteria Check-List (OPCRIT; Reference McGuffin, Farmer and HarveyMcGuffin et al, 1991). Potentially suitable high-risk subjects were approached and interviewed with the Present State Examination (PSE; Reference Wing, Cooper and SartoriusWing et al, 1974) and Schedule for Affective Disorders and Schizophrenia—Lifetime version (SADS—L; Reference Endicott and SpitzerEndicott & Spitzer, 1978) if they agreed to participate. Detailed clinical, neuropsychological and brain imaging assessments were made at intake and have been repeated at 18-month intervals. Subjects have been encouraged to remain in regular informal contact with the investigators, and their general practitioners have been contacted at regular intervals in an attempt to ensure that any psychotic symptoms are detected. In total, 229 high-risk individuals have been identified and approached, of whom 162 have thus far provided some data. Some of these subjects have experienced psychotic symptoms as elicited with the PSE (Reference Johnstone, Abukmeil and ByrneJohnstone et al, 2000).

Two control groups have also been recruited (Reference Hodges, Byrne and GrantHodges et al, 1999; Reference Johnstone, Abukmeil and ByrneJohnstone et al, 2000). A group of 37 age-matched patients experiencing their first episode of schizophrenia have been identified from admissions to the Royal Edinburgh Hospital and associated psychiatric hospitals in Lothian region, Scotland. These subjects have no known family history of schizophrenia in their first- or second-degree relatives. Their case notes have also been reviewed with OPCRIT and the patients have also been examined with the PSE and SADS—L. A group of 36 healthy controls, also without any family history of psychosis, have been recruited from the same areas of the country as the high-risk subjects.

Rating scales

We obtained ratings on the Neurological Evaluation Scale (NES; Reference Buchanan and HeinrichsBuchanan & Heinrichs, 1989) and the Waldrop Scale (Reference Waldrop and HalversonWaldrop & Halverson, 1971) in 152 high-risk individuals, 30 first-episode patients and 35 healthy controls. Ratings were completed as one of the initial clinical assessments by A.H., R.A.C. and S.M.L. The Waldrop Scale consists of 18 items, in six categories, which are assessed with reference to descriptive anchors and scored (generally 0-2) with a system that is weighted according to abnormalities found in patients with schizophrenia. Two items (head circumference and the binocular distance) are measured with a tape and then assigned to categories by reference to the control mean (as 1.0 or 1.5 standard deviations greater than the control mean). Individual item and total scores are generally calculated. The NES was designed to standardise the neurological assessment of patients with schizophrenia and consists of 26 discrete items, of which 14 are tested and scored for the right and left sides of the body. The instrument includes representative items from three main functional areas of interest — sensory integration, motor coordination and sequencing of complex motor acts — as well as assessment of cerebral dominance, frontal release signs and eye movements. The items are presented in a fixed order and scored on a three-point scale (0, no abnormality; 1, mild but definite impairment; 2, marked impairment) apart from ‘suck’ and ‘snout’ reflexes which are scored 0 or 2. Descriptive anchors are provided to facilitate standardised judgments. Total scores are calculated for each of the three domains above, the remaining tests and overall. In this study, we did not measure cerebral dominance with the NES or include this item in these totals (‘other test’ and overall totals), as this was assessed separately in the Edinburgh High-Risk Study.

Genetic liability

The genetic liability of high-risk individuals to psychosis was measured in two ways. The categorical method simply assigned high-risk individuals to groups of those with only second-degree relatives affected, those with one first- and other second-degree relatives, and those with two or more first-degree relatives with schizophrenia. A continuous measure was also developed for the Edinburgh High-Risk Study by Pak Sham, Reader in Statistical Genetics at the Institute of Psychiatry. A detailed account of the rationale and calculation of the measure is given in Lawrie et al (Reference Lawrie, Whalley and Abukmeil2001). Briefly, we assumed a multifactorial polygenic model of schizophrenia, with a heritability (squared) of 0.7, to generate a continuous but bimodal measure of genetic loading for each individual based on the expected liabilities of all family members.

Statistical analyses

Between-group comparisons of the mean ranks of total scores were made with Kruskal—Wallis analysis of variance (ANOVA) and post hoc Mann—Whitney U-tests. For the within high-risk by symptoms analysis, high-risk individuals were dichotomised into those with any psychotic symptoms elicited on the PSE versus those with only neurotic symptoms or no symptoms at all — see Johnstone et al (Reference Johnstone, Abukmeil and Byrne2000) for further details — and total scores compared with the Mann—Whitney U-test. The analyses of genetic liability were conducted by comparing mean ranks with the Kruskal—Wallis ANOVA and post hoc Mann—Whitney U-tests (for the categorical measure) and by calculating Spearman rank correlation coefficients (for the continuous measure). Finally, we conducted exploratory analyses for each individual test item with χ² tests and Kruskal—Wallis ANOVA as appropriate.

Main results

Our findings support the hypothesis that rates of MPAs differ between high-risk subjects and controls, but these did not differ between subjects at high-risk and those with schizophrenia or by symptoms within the high-risk group. The exploratory findings of large heads in males with schizophrenia and hypertelorism in the patients and high-risk subjects must be interpreted cautiously. Our principal genetic analysis, and two post hoc analyses, found MPAs were, if anything, most common in those with the least genetic liability. These results suggest that MPAs, perhaps particularly of the skull, are trait markers for schizophrenia; but that they are not mediated by the gene(s) for schizophrenia and are not directly related to the development of (positive) psychotic symptoms.

Our NES results showed significant between-group differences in total scores, two domains and three individual tests; although only the sensory integration domain total scores differed between high-risk and control subjects. Only one of 25 comparisons was different between high-risk subjects with and those without psychotic symptoms and no genetic associations were evident. We therefore interpret these findings as evidence that sensory integration abnormalities are non-specific markers of neurodevelopmental abnormality, which cannot be attributed to antipsychotic medication. In general, however, neurological ‘soft signs’ could be a consequence of schizophrenia or its treatment.

Strengths and limitations of this report

Studies of individuals at high risk of developing schizophrenia have the potential to tease out genetic effects in the aetiology of schizophrenia. In our high-risk study, this has been enhanced by the development and use of a measure of genetic liability, and this has been particularly illuminating in respect of neuropsychological (Byrne et al, 1999) and structural variables (Reference Lawrie, Whalley and AbukmeilLawrie et al, 2001). Large high-risk studies which follow-up sufficient numbers of those who become ill also have the ability to distinguish the causes and consequences of psychosis, in particular the associations of symptoms and the effects of antipsychotic medication. We divided our high-risk subjects by the presence or absence of positive psychotic symptoms, mainly sensory distortions and partial delusions, some of which were transient (Reference Johnstone, Abukmeil and ByrneJohnstone et al, 2000) and these results must therefore be regarded as preliminary, particularly as high-risk subjects who develop symptoms at a relatively early age may differ from those who develop them later. The fact that all the high-risk subjects were unmedicated at the time of assessment and indeed had never received antipsychotic medication is, however, a substantial advantage in the interpretation of our results. The very small number of our subjects who have thus far developed psychotic illness means that statistical analysis of the predictors of psychosis would be premature. It would, of course, also be interesting to examine the rates of neurological abnormalities after the onset of psychosis, ideally before and after medication has been prescribed.

The main limitation of the present study is that the raters did not conduct reliability analyses, as we did not wish to risk non-participation in other parts of the study by asking subjects to comply with multiple examinations. The ratings were not blind to group membership but were generally blind to symptom status in the high-risk group as different raters did either all the interviews or all the other clinical assessments on particular days. Any lack of blinding would generally tend to overestimate significant differences whereas any reliability problems may have introduced measurement ‘noise’ that could have obscured some associations. Clearly, however, the former potential problem did not generate many possible false positive differences between high-risk subjects with and without psychotic symptoms and the latter potential problem did not obscure all the possible differences between the three groups of subjects. Indeed, we have subsequently analysed the distance between the eyes in all subjects on their magnetic resonance imaging brain scans, with a completely blind and reliable technique, which confirms the overall differences (although the high-risk group no longer differed from the schizophrenia group) and the lack of differences within the high-risk group by symptoms or genetic liability (Reference Boyes, Whalley and LawrieBoyes et al, 2001).

Replications and refutations

The overall differences we report on the Waldrop Scale are in keeping with the original results of Waldrop & Halverson (Reference Waldrop and Halverson1971) and most other previous investigations (Reference McNeil, Cantor-Graae and IsmailMcNeil et al, 2000). There is, however, only limited support for our exploratory individual item findings. Waldrop & Halverson (Reference Waldrop and Halverson1971) reported that large head circumference and hypertelorism were found in patients with schizophrenia more frequently than in controls, but Green et al (Reference Green, Satz and Gaier1989) reported abnormalities of head circumference only in female patients, and Lane et al (Reference Lane, Kinsella and Murphy1997) found a widened skull base but a reduced binocular diameter in patients. Others have reported an enlarged head circumference in schizophrenia, but a meta-analysis of 10 studies published by 1995 found only a non-significant tendency to larger extracranial size in schizophrenia (Reference Ward, Friedman and WiseWard et al, 1996).

The lack of positive associations between the Waldrop Scale and psychotic symptoms or genetic liability are also consistent with the existing literature. We are not aware of any studies that have linked Waldrop abnormalities to positive symptoms, but there are at least two that do not find such an association (Reference Lohr and FlynnLohr & Flynn, 1993; Reference McGrath, van Os and HoyosMcGrath et al, 1995). There is, however, at least one study linking MPAs and negative symptoms in male patients (Reference O'Callaghan, Buckley and MadiganO'Callaghan et al, 1995), which we did not specifically consider in this study. The fact that we found MPAs in high-risk individuals, from multiply affected families, suggests they could be genetically mediated but the few specific associations we found suggest that even if genetic, they are not attributable to the genes for schizophrenia. Indeed, most other investigators have been unable to find genetic associations of MPAs (Reference Heinrichs and BuchananHeinrichs & Buchanan, 1988; Reference Green, Satz and ChristensonGreen et al, 1994; Reference Griffiths, Sigmundsson and TakeiGriffiths et al, 1998). Ismail et al (Reference Ismail, Cantor-Graae and McNeil1998a ) recently reported increased rates of MPAs in the siblings of patients with schizophrenia, but the precise anomalies differed from those in the patients. The only exception was that both groups tended to have ear abnormalities. We found tendencies to more malformed ears in high-risk subjects than the other groups, but no associations with increasing genetic liability to schizophrenia, on exploratory analyses.

Several investigators have reported an increased rate of ‘soft signs’ in patients with schizophrenia (see reviews by Heinrichs & Buchanan (Reference Heinrichs and Buchanan1988) and Ismail et al (Reference Ismail, Cantor-Graae and McNeil1998b )), but the wide variety of scales used and a general tendency to report total scores rather than presenting individual test data limits the opportunities to compare specific results. It is notable that there are several reports of associations between soft signs and psychotic symptoms, especially negative symptoms and ‘thought disorder’ (Reference Heinrichs and BuchananHeinrichs & Buchanan, 1988; Reference King, Wilson and CooperKing et al, 1991; Reference Arango, Kirkpatrick and BuchananArango et al, 2000). The only difference we found between high-risk subjects with and without (positive) psychotic symptoms was on the glabellar reflex test and this is likely to be a chance finding. Very few of our subjects had negative symptoms or thought disorder, and these could of course adversely affect test performance because of a lack of comprehension of sometimes quite involved instructions and/or a lack of motivation or attentional problems. It is also generally difficult to distinguish primary negative symptoms and those that are secondary to antipsychotic medication. Our results do not resolve these issues, but the findings that sensory integration abnormalities were more evident in (unmedicated) high-risk than in control subjects does at least suggest that ‘soft signs’ in schizophrenia are not entirely attributable to these potential confounders.

The relationship between a family history of schizophrenia and soft signs is also complex. Most studies of patients with schizophrenia have not found such an association (Reference Kolakowska, Williams and JamborKolakowska et al, 1985; Reference Heinrichs and BuchananHeinrichs & Buchanan, 1988; Reference King, Wilson and CooperKing et al, 1991), but the problems with reliably ascertaining family history mean that studies of siblings and high-risk subjects are more reliably informative. Siblings of people with schizophrenia have consistently been shown to have rates of soft signs somewhere between those observed in controls and their affected siblings (Reference Heinrichs and BuchananHeinrichs & Buchanan, 1988; Reference Rossi, De Cataldo and Di MicheleRossi et al, 1990; Reference Ismail, Cantor-Graae and McNeilIsmail et al, 1998b ; Reference Chen, Chen and MakChen et al, 2000). Offspring at high risk have also been shown to have higher rates of sensory—perceptual and motor coordination signs (Reference Marcus, Hans and LewowMarcus et al, 1985). None of these studies has been able to separate specific from non-specific genetic or environmental effects. Our findings are broadly in keeping with these reports, but extend them by showing no association between genetic liability to schizophrenia and neurological soft signs.

Implications

We conclude that MPAs and at least some neurological ‘soft signs’ are over-represented in high-risk samples and patients with schizophrenia, but are not associated with the development of positive psychotic symptoms or genetic liability to schizophrenia. Rather, they probably represent subtle developmental abnormalities that are found in a range of brain disorders. This is in keeping with observations that MPAs are frequently observed in several neuropsychiatric conditions (Reference McNeil, Cantor-Graae and IsmailMcNeil et al, 2000) and that ‘soft signs’ are also evident in patients with other psychoses, affective disorders and substance misuse (Reference Woods, Kinney and Yurgelun-ToddWoods et al, 1986; Reference Lohr and FlynnLohr & Flynn, 1993; Reference McGrath, van Os and HoyosMcGrath et al, 1995).

This is not to say that studying MPAs and soft signs is uninformative, as they can clearly give pointers to the timing and processes of neurodevelopmental deviance in schizophrenia and other disorders. There is, for example, a close relationship between cerebral and craniofacial development (Reference Diewert, Lozanoff and ChoyDiewert et al, 1993). It is possible that these anomalies reflect common underlying factors that are more directly related to the development of schizophrenia. It is even possible that the high-risk subjects in our study may show a more specific relationship between these abnormalities and symptoms as and when a greater number develop frank psychosis, particularly negative symptoms.

Some of our exploratory findings suggest particular avenues for potentially fruitful further study — although these must be viewed cautiously given the large number of statistical tests performed and the lack of strong support for them in the literature. The hypertelorism and head circumference findings suggest that high-risk subjects with these anomalies may be at particular risk of developing schizophrenia, perhaps because they have been exposed to more than one genetic and/or environmental risk factor. Hypertelorism probably reflects abnormal development of the face, which can arise for a number of reasons (Reference Diewert, Lozanoff and ChoyDiewert et al, 1993). Alternatively, relatively distinct disease processes may be involved, that is, some patients with schizophrenia have large heads and others have small heads related to different causal processes. For example, a small head at birth suggests a prenatal (second trimester) developmental abnormality; large heads at birth could reflect arrested hydrocephalus, whereas a large head in adulthood could be due to skull overgrowth secondary to early childhood head trauma or the pleiotropic expression of a developmental gene as in fragile-X syndrome. Finally, it is conceivable that families with facial abnormalities and schizophrenia form a distinct subgroup with variants of velo-cardio-facial or similar syndromes.

CLINICAL IMPLICATIONS

• ▪ Minor physical anomalies (MPAs) and at least some neurological ‘soft signs’ may be trait markers for schizophrenia.

• ▪ ‘Soft signs’ are not simply attributable to the effects of antipsychotic medication.

• ▪ MPAs are non-specific markers of neurodevelopmental deviance.

LIMITATIONS

• ▪ We did not conduct formal assessments of interrater reliability.

• ▪ These results may not be generalisable to all cases of schizophrenia.

• ▪ The associations of individual psychotic symptoms in high-risk subjects may differ from the associations of the constellation of symptoms required for a diagnosis of schizophrenia.