Magnetic resonance imaging

Magnetic resonance imaging (MRI) has established that subtle but widespread abnormalities of cortical and subcortical brain structure are present in schizophrenia on a population basis (Reference Lawrie and AbukmeilLawrie & Abukmeil, 1998; Reference Wright, Rabe-Hesketh and WoodruffWright et al, 2000; Reference Shenton, Dickey and FruminShenton et al, 2001; Reference Honea, Crow and PassinghamHonea et al, 2005). A related technique, diffusion tensor imaging (DTI), is a relatively new modality that can examine, through fractional anisotropy, the microstructure of white matter and, through fibre tractography, aspects of neuronal connectivity in the brain. The majority of a limited number of DTI studies to date (Reference Kanaan, Kim and KaufmannKanaan et al, 2005) have indicated reductions in white matter integrity which supports a broader interpretation of structural brain pathology in schizophrenia to involve cortical disconnectivity (Reference Stephan, Baldeweg and FristonStephan et al, 2006).

Cross-sectional relationship to outcome

Studies have sought to relate structural brain pathology on MRI to aspects of outcome on a cross-sectional basis, usually by subdividing patients retrospectively into ‘good’ v. ‘poor’ outcome groups and comparing them on various MRI measures at a single index assessment. For example, while patients of whatever outcome evidenced smaller thalamic volume relative to controls, those with poor outcome (i.e. hospitalised for more than 50% of the total duration of illness and continuously hospitalised over the past 3 years) evidenced enlargement of the lateral and third ventricles and reduced overall grey matter volume, particularly in the prefrontal cortex, whereas those hospitalised for less than 10% of total duration of illness and not hospitalised over the past year did not (Reference Staal, Hulshoff and SchnackStaal et al, 2001). Similarly, patients of whatever outcome evidenced an overall reduction in grey matter volume relative to controls, particularly in the frontal and temporal lobes, but those with poor outcome, ‘Kraepelinian’ schizophrenia (i.e. continuously hospitalised or completely dependent on others for basic needs, unemployed, with severe negative symptoms and severe thought disorder) evidenced reduced grey matter volume in the temporal and occipital but not in the frontal lobes relative to their good-outcome (i.e. non-‘Kraepelinian’) counterparts (Reference Mitelman, Shihabuddin and BrickmanMitelman et al, 2003).

However, such studies illustrate how the indirect nature and inconsistency of the relationship between MRI measures and retrospective, dichotomous indices of outcome, on a population basis, precludes the generation of a predictive model in relation to the individual patient. Such models can only be generated in longitudinal studies.

Longitudinal relationship to outcome over chronic illness

The putative utility of MRI measures as indices of outcome is associated with the enduring debate on the extent to which structural brain pathology in schizophrenia does or does not progress. The weight of evidence from longitudinal studies beginning at various stages of illness (Reference Pantelis, Yucel and WoodPantelis et al, 2005), together with incisive, pseudo-longitudinal analyses of cross-sectional studies (Reference Woods, Ward and JohnsonWoods et al, 2005), now suggests a small but significant acceleration in the loss of cortical grey matter volume with associated enlargement of cerebrospinal fluid (CSF) spaces in the long term. To the extent that one can generalise from very early-onset schizophrenia in children to the more typical presentation in young adulthood, there is initial evidence for dynamic tissue loss that progresses to frontal, but less so to cingulate and temporal cortex, in ‘waves’ along the anterior–posterior and dorsal–ventral axes (Reference Vidal, Rapoport and HayashiVidal et al, 2006).

Studies have related such longitudinal changes to aspects of outcome, usually in terms of MRI measures and assessments of clinical course made on two occasions at the beginning and end of a given period of follow-up. For example, over a mean interval of 3.6 years (range 0.6–7.5), patients who had been ill for a mean of 15.3 years (range 2.2–26.5) evidenced greater decline in cortical grey matter and greater increase in both frontal and temporal cortical sulcal and ventricular CSF volumes relative to controls. The rate of expansion in frontal sulci was associated with higher overall positive symptom scores and a longer percentage of time spent in hospital over follow-up; the rate of decrease in grey matter and of sulcal expansion in prefrontal cortex was associated with higher overall negative symptom scores and longer percentage of time spent in hospital; and the rate of decrease in grey matter and of sulcal expansion in temporal cortex was associated with higher overall negative symptom scores (Reference Mathalon, Sullivan and LimMathalon et al, 2001).

However, although such studies support the presence of some subtle but poorly understood neuroprogressive process in schizophrenia, they involve people with differing durations of chronic illness at the start of variable periods of longitudinal assessment who are assessed using a limited range of outcome measures, on a population basis. Thus, they are unable to generate a predictive model in relation to individual patients. Such models can only be generated in prospective studies from the first psychotic episode.

Prospective relationship to outcome from the first episode

It is now clear that some aspects of structural brain pathology evident on MRI in chronic schizophrenia, particularly reduction in whole brain and hippocampal volumes, together with enlargement of the third and lateral ventricles, are present at the time of the first psychotic episode and therefore pre-date onset of diagnostic symptoms (Reference Steen, Mull and McClureSteen et al, 2006; Reference Vita, De Peri and SilenziVita et al, 2006). Diffusion tensor imaging also indicates lower fractional anisotropy, consistent with white matter abnormalities at the first episode (Reference Szeszko, Ardekani and AshtariSzeszko et al, 2005). Thereafter, as considered above in the context of chronic illness, there is a small but significant acceleration in loss of cortical grey matter volume with associated enlargement of CSF spaces (Reference Pantelis, Yucel and WoodPantelis et al, 2005; Reference Whitford, Grieve and FarrowWhitford et al, 2006). The ‘anchor’ event of the first psychotic episode provides a frame of reference from which any relationships between MRI parameters and long-term outcome can be explored prospectively.

Studies have related such longitudinal change to prospective evaluation of outcome, usually in terms of MRI measures and assessments of clinical course made on two occasions: at the first episode and at variable periods of follow-up. For example, in a prospective study over 3 years, extent of ventricular enlargement during this period was associated with poor outcome as dichotomised in terms of remission of positive symptoms; extent of reduction in frontal lobe white matter volume and of increase in frontal lobe sulcal CSF volume were associated with greater negative symptom severity; extent of reduction in frontal lobe grey matter volume was associated with poorer executive functioning over follow-up (Reference Ho, Andreasen and NopoulosHo et al, 2003).

Although this study addressed relationships between longitudinal changes in MRI parameters and outcome measures, a critical question is whether any cross-sectional MRI measure made at the first episode is predictive of subsequent outcome as assessed prospectively. In a prospective study over 2 years, total brain volume and volumes of cortical grey and white matter, third and lateral ventricles and cerebellum at the first or second episode failed to predict outcome in terms of positive or negative symptoms, social disability and need for care (Reference van Haren, Chan and Polvan Haren et al, 2003). Similarly, in a prospective study over 5 years, smaller temporal lobe grey matter volume at the first episode was associated with persistence of hallucinations; however, initial temporal and frontal lobe tissue and sulcal and ventricular CSF volumes were unrelated to negative symptoms, extent of hospitalisation or psychosocial outcome (quality of relationships, sexual activity, recreation and work performance) over follow-up (Reference Milev, Ho and ArndtMilev et al, 2003). In a further prospective study over 2 years, decreasing volume of the dorsolateral prefrontal cortex at the first episode was associated with poorer functional outcome, in terms of social and employment indices, at 1 but not at 2 years (Reference Prasad, Sahni and RohmPrasad et al, 2005).

However, although such studies further elaborate the presence of some subtle but poorly understood neuroprogressive process in schizophrenia, they indicate that MRI measures made at and following the first psychotic episode have only limited capacity to predict outcome on a patient population basis. Thus, they are unable to generate a predictive model in relation to individual patients. One potential confound in all such studies is an effect of long-term antipsychotic treatment on brain tissue volumes. Magnetic resonance imaging studies have indicated that volumes of the basal ganglia and pituitary are increased following exposure to typical but not atypical antipsychotics (Reference Lieberman, Toffefson and CharlesLieberman et al, 2005; Reference Pariante, Dazzan and DanesePariante et al, 2005). More extensive studies have indicated exposure to typical antipsychotics to be associated with some reduction in cortical grey matter volume, whereas atypical agents can be associated with some increase in this volume (Reference Dazzan, Morgan and OrrDazzan et al, 2005; Reference Garver, Holcomb and ChristensenGarver et al, 2005; Reference Lieberman, Toffefson and CharlesLieberman et al, 2005). Given the practical difficulties in conducting MRI studies during the first episode in a person with no antipsychotic treatment, especially in a clinical setting, these effects could clearly confound the search for relationships between MRI measures and outcome.

Prospective relationship to transition from high-risk/prodrome to first episode

Over the past several years there has emerged a yet earlier domain of outcome: that of transition v. non-transition from high-risk status (defined usually in terms of affected family member(s), attenuated or transient psychotic symptoms and functional decline) to first-episode psychosis.

In a novel and important MRI study over a prospective period of at least 1 year (Reference Pantelis, Velakoulis and McGorryPantelis et al, 2003), those high-risk participants who developed first-episode psychosis were characterised by a reduction in parahippocampal, fusiform and orbito-frontal cortex, and in cingulate gyri; for those who did not develop a first psychotic episode, no such longitudinal changes were evident. On addressing the critical question of whether any cross-sectional MRI measure made in the high-risk state is predictive of outcome as assessed prospectively, transition to a first psychotic episode was associated with reduced grey matter volume in the medial and lateral temporal, inferior frontal and cingulate cortex.

In a related study over a prospective period of 5 years, those high-risk participants who received two MRI scans at an interval of 2 years and subsequently received a diagnosis of schizophrenia were distinguished from those who did not receive this diagnosis by reduction in grey matter density in temporal gyrus, uncus and cerebellum between those two scans; cross-sectional MRI measures made in the high-risk state did not differ between those who did and did not go on to receive a diagnosis of schizophrenia (Reference Job, Whalley and JohnstoneJob et al, 2005). However, transition to psychosis, not just to schizophrenia but also particularly to psychotic depression, is associated with larger pituitary volume in the high-risk state, suggesting stress-related activation of the hypothalamic–pituitary–adrenal axis (Reference Garner, Pariante and WoodGarner et al, 2005). However, it is far from clear whether such findings, made on a patient population basis, allow generation of a predictive model in relation to individual patients.

Functional magnetic resonance imaging

Functional magnetic resonance imaging (fMRI) accesses regional neuronal activation in terms of change in blood oxygenation. Although a variety of cross-sectional fMRI findings in schizophrenia have been reported (Reference Tost, Ende and RufTost et al, 2005), systematic, prospective studies from the high-risk state or the first psychotic episode in relation to outcome are in their infancy.

A recent cross-sectional study has examined frontal and cingulate cortex, thalamic and basal ganglia activation during executive processing in those at high risk and in both early-phase and patients with chronic illness in comparison with controls (Reference Morey, Inan and MitchellMorey et al, 2005). The findings indicate that prefrontal function begins to decline before the emergence of diagnostic symptoms and impairment in frontostriatal function is evident thereafter. Recently, a prospective study over a period of 5 years has examined activation during a sentence completion task (Reference Whalley, Simonotto and MoorheadWhalley et al, 2006). Cross-sectional fMRI measures made in the high-risk state indicated that those who went on to receive a diagnosis of schizophrenia were distinguished from those who did not by decreased activation of the anterior cingulate, increased activation of the parietal lobe and smaller increases in activation with increasing task difficulty. However, only four high-risk participants evidenced transition to schizophrenia over the follow-up period (Reference Whalley, Simonotto and MoorheadWhalley et al, 2006). The elaboration of such fMRI studies in those at high risk and first-episode patients on a prospective basis has the potential to provide important additional information on the prediction of long-term outcome. However, possible confounding effects of antipsychotic drugs on fMRI measures remain a cause for concern (Reference Davis, Jeste and EylerDavis et al, 2005).

Magnetic resonance spectroscopy

Magnetic resonance spectroscopy (MRS) also accesses functional processes, primarily the concentration of various brain metabolites, to provide more detailed information on regional cellular mechanisms.

Although a variety of cross-sectional MRS findings in schizophrenia have been reported (Reference Keshavan, Berger and ZipurskyKeshavan et al, 2005), only one study to date has systematically applied this technique prospectively from the first psychotic episode in relation to outcome. Over a prospective period of 1.5 years, a lower N-acetylaspartate/creatine ratio, a putative index of neuronal integrity, at the first episode was associated with poorer outcome in terms of global assessment of functioning, social and occupational functioning and number of hospital admissions (Reference Wood, Berger and LambertWood et al, 2006). As for fMRI, the extension of MRS studies to those at high risk and first-episode patients on a prospective basis, to include assessments beyond transition to psychosis, has the potential to provide important additional information on prediction of long-term outcome. Similarly, as for fMRI, possible confounding effects of antipsychotic drugs on MRS measures remain a cause for concern (Reference Davis, Jeste and EylerDavis et al, 2005).

Emission tomography

Positron emission tomography (PET) and single-photon emission computed tomography (SPECT) are related techniques that access functional processes such as receptor availability, drug–receptor occupancies, transmitter biosynthesis/catabolism and cerebral metabolic activity. Although a variety of cross-sectional PET/SPECT findings in schizophrenia have been reported, only a few studies have systematically applied these techniques prospectively from the high-risk state or the first psychotic episode in relation to outcome. These relate primarily to contemporary versions of the long-standing dopamine hyperactivity model of psychosis, as recently elaborated (Reference Kapur, Mizrahi and LiKapur et al, 2005; Reference Seeman, Schwarz and ChenSeeman et al, 2006).

Over a prospective period of 2 years, higher SPECT binding of ¹²³I-iodobenzamide to striatal D₂ dopamine receptors at the first psychotic episode was associated with poorer social and occupational outcome among those who attained a follow-up diagnosis of schizophrenia but not among those who retained a diagnosis of schizophreniform disorder (Reference Corripio, Perez and CatafauCorripio et al, 2006). Recently, a PET study of striatal ¹⁸F-DOPA uptake has indicated dopamine overactivity in people in the high-risk state. These people are being followed prospectively to determine whether those who go on to evidence a first psychotic episode have raised ¹⁸F-DOPA uptake compared with those who do not (Reference Howes, Montgomery and AsselinHowes et al, 2006). Elaboration of such prospective PET/SPECT studies of those at high risk to include assessment of outcome beyond transition to psychosis has the potential to provide important additional information on prediction of long-term outcome. As for fMRI and MRS, possible confounding effects of antipsychotic drugs on PET/SPECT measures remain a cause for concern (Reference Davis, Jeste and EylerDavis et al, 2005).

Electroencephalography

Although electroencephalography (EEG) has a history in biological psychiatry that long pre-dates the emergence of MRI, fMRI, MRS and PET/SPECT techniques, it has failed to match their impact. The recent introduction of techniques such as gamma synchrony has contributed to a new wave of incisive studies.

For example, a cross-sectional study in first-episode schizophrenia has reported decreased magnitude and delayed latency for frontal gamma 1 but not gamma 2 synchrony time-locked to target auditory stimuli, indicating disturbance in connectivity of neural activity in early sensory response to task-relevant stimuli, in a manner that may be modulated by antipsychotic drugs (Reference Symond, Harris and GordonSymond et al, 2005). The extension of such EEG studies to those at high risk and first-episode patients on a prospective basis has the potential to provide important additional information on prediction of long-term outcome.

Neurological soft signs

Neurological soft signs are non-localising abnormalities that cannot be related to impairment in a specific brain region and are not part of a well-defined neurological syndrome. They constitute evidence for otherwise unspecified brain dysfunction and have been shown consistently to occur to excess in schizophrenia (Reference Bombin, Arango and BuchananBombin et al, 2005).

Studies have indicated that the extent of neurological soft signs at the first episode has little relationship to long-term outcome, for example global or occupational functioning (Reference Bombin, Arango and BuchananBombin et al, 2005). Recently, in a study over a prospective period of 4 years from the first episode, improvement in neurological soft signs score was associated with better overall outcome over the same period (Reference Whitty, Clarke and McTigueWhitty et al, 2006). However, in a study over a prospective period of 1 year, extent of neurological soft signs at the first episode failed to predict outcome in terms of psychopathology or rate of relapse, defined as hospitalisation or unscheduled visit due to exacerbation, but did predict emergence of tardive dyskinesia (Reference Emsley, Turner and OosthuizenEmsley et al, 2005). Similarly, in a study over a prospective period of 3 years, extent of neurological soft signs at the first episode failed to predict outcome in terms of relapse or occupational functioning (Reference Chen, Hui and ChanChen et al, 2005). Thus, the findings to date do not indicate that neurological soft signs exert material prediction of long-term outcome.

Anthropometrics of craniofacial dysmorphogenesis

Aspects of dysmorphogenesis, particularly of craniofacial regions that bear the most intimate embryological relationship with early brain development, can be measured using classic anthropometric techniques (Reference Lane, Kinsella and MurphyLane et al, 1997). However, in a recent prospective study over a period of 5 years, cross-sectional measurement of hypertelorism made in the high-risk state did not distinguish those who went on to receive a diagnosis of schizophrenia from those who did not (Reference Johnstone, Ebmeier and MillerJohnstone et al, 2005).

Three-dimensional surface imaging of craniofacial dysmorphogenesis

It is now possible to apply three-dimensional surface imaging technology and geometric morphometrics to the quantitative measurement of facial dysmorphology (Reference Hennessy, McLearie and KinsellaHennessy et al, 2005). Such techniques, which have been shown recently to distinguish people with schizophrenia from controls (Reference Hennessy, Baldwin and BrowneHennessy et al, 2007), might have greater potential for addressing any relationship between dysmorphogenesis and long-term outcome in schizophrenia.

Dermatoglyphics

Another index of dysmorphogenesis is dermatoglyphics; for example, a-b ridge count, a quantitative dermatoglyphic measure of the palm, is reduced in schizophrenia (Reference Bramon, Walshe and McDonaldBramon et al, 2005). However, in a recent prospective study over a period of 5 years, cross-sectional measurement of dermatoglyphics made in the high-risk state did not distinguish those who went on to receive a diagnosis of schizophrenia from those who did not (Reference Johnstone, Ebmeier and MillerJohnstone et al, 2005).

Postnatal MRI

Although brain MRI in the immediate postnatal period may have some utility in predicting neurodevelopmental outcome at 2 years (Reference Woodward, Anderson and AustinWoodward et al, 2006), any ability to predict schizophrenia as an outcome in young adulthood remains unexplored.