Participants

Participants were recruited as part of a large epidemiological study of first-episode psychosis (Reference Dazzan, Morgan and OrrDazzan et al, 2004). A total of 281 people met inclusion criteria for a functional psychotic illness (DSM–IV 295–298, psychotic codings; American Psychiatric Association, 1994); 90 declined to participate; of the remaining 191, 115 consented to an MRI scan. These 115 patients were a mean of 6 years younger (mean age 27.9 years, s.d.=8.4 v. 33.7, s.d.=12.3, t=3.5, P=0.001) and had a higher proportion of White British members (36% v. 18%, χ²=6.95, P=0.008) but were otherwise comparable to the total sample in terms of gender, years of education, diagnosis and duration of illness. There were no significant group differences in demographic or clinical features between those who did and did not consent to scanning.

Of those who consented to MRI, 10 terminated scanning before image acquisition could be completed and 15 were excluded owing to participant motion (n=13) or radiological abnormalities (1 congenital hydrocephalus, 1 subarachnoid cyst). Data from 90 participants were thus available for image analysis.

Clinical measures

Participants were assessed using the Schedule for Clinical Assessment in Neuropsychiatry (SCAN; World Health Organization, 1994) interview. Diagnostic codes were assigned according to DSM–IV criteria in consensus meetings with senior clinicians. A total symptom score was obtained by summing the individual symptom item scores according to Wing & Sturt's (1978) procedure for the Present State Examination (PSE). Premorbid IQ was assessed using the National Adult Reading Test (NART; Reference Nelson and WillisonNelson & Willison, 1991) and handedness using the Annett Hand Preference Questionnaire (Reference AnnettAnnett, 1970).

Prodrome

The prodrome was defined as the period in weeks from the time of first definite change in behavioural, psychological or emotional functioning to the onset of the first psychotic episode. The key feature of this criterion was the presence of a fundamental change in function that persisted from its onset until the time of first episode of psychosis (e.g. non-attendance at work/college with no clear explanation for change in behaviour). In line with previous studies (Reference Craig, Bromet and FennigCraig et al, 2000), onset of psychosis was defined as the presence for 1 week or more of one of the following psychotic symptoms (as defined in the SCAN): delusions; hallucinations; marked thought disorder; marked psychomotor disorder (other than simple retardation or acceleration); and bizarre, grossly inappropriate and/or disorganised behaviour with a marked deterioration in function. Data relating to date of first behavioural change and date of onset of psychosis were collated from interviews with the participant and a close relative, and clinical notes using the Personal and Psychiatric History Schedule (PPHS; Reference Jablensky, Sartorius and ErnbergJablensky et al, 1992; World Health Organization, 1994). Interrater reliability was assessed for those involved in rating (C.M. and J.M.L.) by each independently rating these measures on a random subset of 50 participants from the original sample. Reliability was judged as satisfactory (inter-class correlation r=0.903). Duration of untreated psychosis was measured and defined as the period in weeks from the onset of frank psychotic phenomena to first contact with statutory mental health services. Duration of untreated illness was defined as the sum of duration of prodrome plus the duration of untreated psychosis.

MRI protocol

Magnetic resonance images were acquired with a GE Signa 1.5 T system (GE Medical Systems, Milwaukee, USA) at the Maudsley Hospital (London, UK). Contiguous, high-resolution 3-mm thick, interleaved dual-echo images were acquired in the coronal plane, providing coverage of the entire brain. Proton density- and T2-weighted image acquisition was almost simultaneous: time to echo (TE1) 20 ms, (TE2) 85 ms; time to resolution (TR) 4000 ms; 8-echo train length; matrix size 256 × 192; field of view 22 cm×16.5 cm).

Images were processed using SBAMM voxel-based semi-automated methods (Reference Bullmore, Suckling and OvermeyerBullmore et al, 1999; Reference Suckling, Sigmundsson and GreenwoodSuckling et al, 1999). In brief, extracerebral tissues were removed using an automated algorithm and manual editing. Intracerebral tissue voxels were categorised into grey matter, white matter, cerebrospinal fluid (CSF), or dura/vasculature using a modified fuzzy clustering algorithm (Reference Suckling, Sigmundsson and GreenwoodSuckling et al, 1999). A template image was constructed in standard space from proton-density images of six healthy controls using the AFNI program (Reference Dazzan, Morgan and OrrDazzan et al, 2004). Each participant's proton-density weighted image was then registered onto the template with a nine-parameter affine registration, minimising between-level difference between images. The derived mapping was applied to the corresponding tissue maps.

Length of prodrome was regressed onto estimated grey matter and white matter volumes at each voxel in standard space. The test statistic calculated was the regression coefficient divided by its standard deviation to generate an effect map. Permutation testing was used to assess statistical significance, and regional relationships were tested at the level of voxel clusters (Reference Bullmore, Suckling and OvermeyerBullmore et al, 1999; Reference Sigmundsson, Suckling and MaierSigmundsson et al, 2001). The statistical threshold for cluster significance was set such that the expected number of false-positive clusters (P value times number of tests) was less than one.

Sample

Socio-demographic and clinical characteristics of the sample are shown in Table 1.

Table 1 Socio-demographic and clinical characteristics of 82 participants

Characteristic
Length of prodrome, weeks
Mean (s.d.)	21 (36)
Median (IQR)	5 (1-27)
Duration of untreated psychosis, weeks
Mean (s.d.)	26 (52)
Median (IQR)	6 (2-22)
Duration of untreated illness, weeks
Mean (s.d.)	46 (65)
Median (IQR)	22 (7-53)
Male gender, n (%)	51 (62)
Right handedness, n (%)1	71 (90)
Diagnosis, n (%)
Schizophrenia	46 (56)
Affective psychoses	26 (31)
Other psychoses	10 (12)
Age, years: mean (s.d.)	27.2 (8)
Education, years: mean (s.d.)	12.8 (2)
Premorbid IQ: mean (s.d.)2	95.9 (16)
Typical antipsychotics, n (%)	38 (46)
Atypical antipsychotics, n (%)3	33 (40)
Drug-free, n (%)	12 (15)
Days on treatment: mean (s.d.)
Typical antipsychotics	62.6 (59)
Atypical antipsychotics	63 (59)
Current dose
Typical antipsychotics, mg in chlorpromazine equivalents: mean (s.d.)	283.5 (233)
Atypical antipsychotics, mg olanzapine: mean (s.d.)	13.4 (5)
Total symptom score (SCAN)	30.2 (16)
Total grey matter volume, cm3: mean (s.d.)	576.2 (60)
Total white matter volume, cm3: mean (s.d.)	469.0 (51)
Total CSF volume, cm3: mean (s.d.)	154.2 (30)

Prodrome

The distribution of prodrome length expressed in weeks was non-Gaussian. A small number of participants (n=8) had an extremely long prodrome (>169 weeks), which was grossly outwith the distribution for the sample as a whole. Logarithmic transformation was not sufficient to convert the data to normal distribution, and therefore these participants were excluded as statistical outliers.

Eighty-two participants remained for subsequent analyses; these were a mean of 5 years younger (mean age 27.1, s.d.=8.0 v. 32.7, s.d.=11.1 years, t=3.8, P=0.002) and had a higher proportion of White British members (38% v. 21%, χ²= 6.24, P=0.01), but were otherwise comparable to the total sample in terms of gender, years of education, diagnosis and duration of illness.

The mean length of prodrome was 21 weeks (s.d.=36, range 1–157, IQR 1=27). The mean duration of untreated psychosis was 26 weeks (s.d.=52, range 1–313, IQR=2–22). The mean duration of untreated illness was 46 weeks (s.d.=64, range 1–344, IQR=7–52). Seventy participants (85%) had received antipsychotic treatment at the time of the MRI scan. The mean duration of antipsychotic treatment was 9 weeks.

Length of prodrome did not vary significantly with gender, age, type of psychotic disorder (schizophrenia, affective psychosis, other), total duration of antipsychotic exposure, illness severity (as indexed by total symptom score on the SCAN) or premorbid IQ (NART; Table 2).

Table 2 Relationship of socio-demographic and clinical variables to prodrome length

Characteristic	Prodrome length Mean (s.d.)
Diagnosis
Schizophrenia (n=46)	22.5 (38)
Affective psychosis (n=26)	16.5 (39)
Other (n=10)	22.8 (18)
Gender
Male (n=51)	16.9 (32)
Female (n=31)	26.8 (43)
Marital Status
Married or steady relationship (n=20)	20.1 (28)
Single, separated or divorced (n=59)	18.6 (37)
Occupational status
Employed (n=35)	21.9 (42)
Unemployed (n=44)	17.9 (28)
Antipsychotic treatment
Typicals (n=24)	21.6 (46)
Atypicals (n=19)	12.4 (21)
Mixed (n=14)	32.9 (46)
None (n=25)	19.1 (30)

Global brain volumes

Length of prodrome was not significantly correlated with the total volume of any of the grey matter, the white matter or the CSF (Table 2).

Regional brain volumes

Grey matter

There were two regions where there was an inverse relationship between length of prodrome and grey matter volume (P=0.002, < 1 false-positive; Table 3; Fig. DS1 in the data supplement to the online version of this paper). The first cluster included the anterior and posterior cingulate gyri bilaterally, and extended locally into the medial frontal gyri bilaterally (Fig. DS1). A second cluster was focused on the left insula, extending into the left inferior frontal gyrus.

Table 3 Regional differences in grey and white matter associated with length of prodrome

Anatomical area	Cluster size voxel number	Location of cluster centre, x, y, z
Gray matter reductions:
R cingulate gyrus, extending to L cingulate gyrus, and R and L medial frontal gyri	641	0.2, 4.3, 38.8
L insula, extending to L inferior frontal gyrus	362	-35.5, 1.9, 2.2
White matter reductions
L superior longitudinal fasciculus, with extension to L cingulum, L corpus callosal and L internal capsule fibres	2133	-24.4, -6.7, 26.2
R superior longitudinal fasciculus, with extension to R cingulum, R corpus callosal and R internal capsule fibres	2729	24.7, 3.1, 24.9
White matter excesses
L acoustic radiation, with involvement of L inferior longitudinal fasciculus and L uncinate fasciculus	182	-34.7, -30.4, 8.9
R inferior longitudinal fasciciulus, extending to R uncinate fasciculus	408	42.5, -26.8, -8.7
R orbital gyral white matter, extending to fusiform white matter	384	32.2, -68.0, -21.6

Grey and white matter abnormalities

This study provides the first data to indicate that the length of time spent in prodromal illness directly correlates with the degree of brain structural abnormalities present at the onset of psychosis. Moreover, this relationship was evident in grey and white matter areas that are typically affected in psychotic disorders, and the polarity of the relationship in these areas (smaller volume with longer prodrome) corresponds to the reduction that is evident in psychotic disorders. Reduced prefrontal, insular and cingulate volume have consistently been reported in neuroimaging studies of schizophrenia (Reference Wright, Rabe-Hesketh and WoodruffWright et al, 2000; Reference Sigmundsson, Suckling and MaierSigmundsson et al, 2001; Reference Job, Whalley and McConnellJob et al, 2002) and, to a lesser extent, bipolar illness (Reference Drevets, Price and SimpsonDrevets et al, 1997; Reference McDonald, Bullmore and ShamMcDonald et al, 2004). Similarly, reduced white matter volume in the superior longitudinal and uncinate fasciculi, and in the cingulum has been described in volumetric MRI studies of schizophrenia (Reference Sigmundsson, Suckling and MaierSigmundsson et al, 2001) and bipolar disorder (Reference Haznedar, Roversi and PallantiHaznedar et al, 2005). Diffusion tensor imaging studies indicate that these tracts are also sites of reduced fractional anisotropy in schizophrenia (interpreted as reflecting reduced white matter integrity), but this technique has been less frequently employed in bipolar disorder (Reference Lim, Hedehus and MoseleyLim et al, 1999; Reference Kubicki, Westin and MaierKubicki et al, 2002). Our white matter findings are of particular interest because the superior longitudinal fasciculi and the cingulum carry connections to and from the prefrontal, cingulate and temporal regions (Reference Catani, Howard and PajevicCatani et al, 2002; Reference Wakana, Jiang and Nagae-PoetscherWakana et al, 2004), where we found grey matter reductions correlated with longer prodrome.

Comparison with high-risk studies

A correlation between the duration of the prodromal phase and the severity of MRI abnormalities at the first episode is consistent with data from prospective MRI studies of individuals at high risk of psychosis during the prodromal period. To date, only two previous studies have examined changes in brain structure at several time points in small numbers of people at high risk for psychosis. Pantelis et al (Reference Pantelis, Velakoulis and McGorry2003) scanned people in the prodromal phase and again after they developed psychosis and found that there was a progressive reduction in grey matter volume in the cingulate, prefrontal and medial temporal cortex. Job et al (Reference Job, Whalley and Johnstone2005) also reported a longitudinal reduction in medial temporal and cerebellar volume during the prodrome, and that these changes occurred within the prodromal phase well in advance of the first episode. Our study provides complementary findings: we have established that there is a correlation between length of prodromal symptoms and severity of neuroanatomical abnormalities. This is important because if these occur secondary to progressive changes caused by subthreshold prodromal symptoms, it is clear that the length of prodromal illness is important in pathogenesis, rather than there being some ‘toxic’ threshold being reached with the onset of frank psychotic symptoms.

However, although our findings are consistent with this hypothesis, we cannot conclude from our cross-sectional study that the brain abnormalities identified represent progressive changes. Indeed, the differences in regional volumes within our participants with first-episode psychosis may already have been present before the prodromal phase, perhaps reflecting factors that vary between patients and differentially affect both brain structure and the form of onset of the disorder. For example, this explanation might apply if those who developed schizophrenia had both a longer prodrome and more extensive MRI abnormalities than those with an affective psychosis, or if people who were highly symptomatic at the first episode had a longer prodrome than those who presented with less florid symptoms. In the present study the length of prodrome was similar for people with schizophrenia, affective or other psychoses, and it was also unrelated to the overall severity of symptoms. Similarly, prodrome duration was unrelated to age, gender, IQ, nature of antipsychotic treatment and duration of treatment prior to scanning. Thus, although an effect of factors other than a progressive process cannot be excluded, we found no evidence to support this. None the less, it is important to acknowledge the possibility that any changes occurring in brain structure need not follow a purely linear course throughout the entire prodrome. Indeed it seems likely that, after a given time, rate of change must diminish or brain volume would decrease to zero. Our interpretation of the findings as representing a linear change over time is both scientifically plausible and parsimonious.

Definition of prodrome

Estimates of prodrome duration in the present study were made retrospectively, which might have limited their accuracy. We sought to minimise inaccuracy by using information from the participants' relatives and friends, as well as from the participants themselves, and by systematically examining all available clinical records. Moreover, in practice a truly prospective measure of prodrome duration is difficult to obtain. Even people who are ascertained before the first episode will have been experiencing prodromal symptoms for some time (Reference Yung, Phillips and YuenYung et al, 2003), such that their date of illness onset still has to be estimated retrospectively. In any event, in the present study it is the relative length of the prodrome in different participants rather than its absolute duration that is relevant, making the precision of the absolute estimate less critical. The average length of prodrome in our sample was shorter than reported in previous studies (Reference Pantelis, Velakoulis and McGorryPantelis et al, 2003), and this might reflect differences in the respective samples examined. Our participants were people with first-episode psychosis ascertained in an epidemiological study, whereas previous estimates have been derived from people with prodromal symptoms referred to specialist clinics (Reference Pantelis, Velakoulis and McGorryPantelis et al, 2003).

Voxel-based methodology

We used voxel-based morphometry for the evaluation of brain structure, measuring relative intensities of grey or white matter values at a given voxel. Thus, observed differences between groups might represent differences in factors affecting tissue intensity such as hydration status, or vascular or metabolic changes. Also, it is always possible that some of the grey and white matter abnormalities that we have demonstrated could be a result of an artefact of the warping of the images into standard (Talairach) space.

We also considered the possibility that brain changes might occur secondary to antipsychotic treatment in our subjects. There was no difference in length of prodrome between participants considered according to their treatment status (typical antipsychotics, atypicals, mixed, or drug-free; Table 2), so medication status is unlikely to explain our findings.

Relationship with duration of untreated psychosis

The duration of untreated psychosis has been regarded as another period when a neurotoxic process may be active (Reference WyattWyatt, 1995). Some studies have reported that the severity of volumetric abnormalities in schizophrenia varies with its duration (Reference Keshavan, Haas and KahnKeshavan et al, 1998), but others have not (Reference Hoff, Sakuma and RaziHoff et al, 2000; Reference Ho, Alicata and WardHo et al, 2003). We have previously examined the volumetric correlates of duration of untreated psychosis in this same sample, and found that a long duration was associated with more extensive reductions in grey matter volume that were limited to a single region (temporal cortex) distinct from those identified in the present study (Reference Lappin, Morgan and MorganLappin et al, 2006). Taken together, these data invoke the question of whether there are dynamic changes occurring across integrated brain networks in a temporal sequence. Thus, changes occurring in the prodromal phase predominate in the frontal regions and white matter, whereas later changes related to the transition from prodrome to psychosis (when the duration of untreated psychosis begins) involve other areas such as the temporal cortex. Studies of longitudinal design with scans at several time points are necessary to investigate this possibility.

Clinical implications

Conventionally, clinical intervention in psychotic disorders begins after the first episode. However, there is increasing interest in the potential benefits of introducing treatment in the prodromal phase (Reference McGorry, Yung and PhillipsMcGorry et al, 2002). Although our findings are consistent with the operation of a progressive process during this period, this requires confirmation through further longitudinal MRI studies of people within the prodrome.