Anatomic Abnormalities of the Anterior Cingulate
Cortex Before Psychosis Onset: An MRI Study of
Alex Fornito, Alison R. Yung, Stephen J. Wood, Lisa J. Phillips, Barnaby Nelson, Sue Cotton,
Dennis Velakoulis, Patrick D. McGorry, Christos Pantelis, and Murat Yücel
but whether such changes are apparent before psychosis onset remains unclear. In this study, we characterized prepsychotic ACC
abnormalities in a sample of individuals at ultra-high-risk (UHR) for psychosis.
did not (UHR-NP; n ? 35), and healthy control subjects (n ? 33).
Results: Relative to control subjects, UHR-P individuals displayed bilateral thinning of a rostral paralimbic ACC region that was negatively
correlated with negative symptoms, whereas UHR-NP individuals displayed a relative thickening of dorsal and rostral limbic areas that was
symptomatology. Subdiagnostic comparisons revealed that changes in the UHR-P group were driven by individuals subsequently diag-
nosed with a schizophrenia spectrum psychosis.
Conclusions: These findings indicate that anatomic abnormalities of the ACC precede psychosis onset and that baseline ACC differences
distinguish between UHR individuals who do and do not subsequently develop frank psychosis. They also indicate that prepsychotic
changes are relatively specific to individuals who develop a schizophrenia spectrum disorder, suggesting they may represent a diagnosti-
cally specific risk marker.
Key Words: Bipolar disorder, depression, limbic system, mania,
pathogenesis of psychotic disorders (1–8), but it remains unclear
whether these abnormalities precede or follow psychosis onset.
Establishing the timing of such neuroanatomic changes is critical
to determine whether they represent a risk marker for illness
onset or a secondary manifestation of the disease process.
To determine whether neuroanatomic differences repre-
sent risk markers, it is necessary to identify and follow-up
individuals at elevated risk for psychosis to characterize
diagnostic outcomes. To our knowledge, only three such MRI
studies incorporating ACC measures have been published.
Two have investigated individuals at ultra-high-risk (UHR) for
psychosis identified using state and trait criteria (9,10). Both
reported reduced ACC gray matter in UHR individuals who
subsequently developed psychosis (UHR-P) compared with
those who did not (UHR-NP). Job and colleagues (11) found
ounting neuropathologic and magnetic resonance im-
aging (MRI) evidence supports a primary role for ab-
normalities of the anterior cingulate cortex (ACC) in the
reduced ACC gray matter in individuals at genetic high risk for
schizophrenia compared with healthy control subjects (11),
but these reductions were no more severe in those high-risk
individuals who eventually developed schizophrenia or sub-
threshold psychotic symptoms (12).
One limitation of these reports is that baseline ACC measures
were not compared with healthy control subjects, making it
unclear whether differences between high-risk people who do
and do not develop psychosis reflect a putative abnormality.
Second, the rates of transition to psychosis have been small (max-
imum n ? 23) (9), precluding the opportunity to investigate
subdiagnostic specificity. A third limitation is the use of voxel-based
morphometry (VBM), which is restricted, by the image resolution
(typically ? 1 mm3), in its capacity to detect relatively subtle
changes. Further, VBM fails to account for the considerable inter-
individual anatomic variability of the ACC, particularly with respect
to the paracingulate sulcus (PCS), a highly variable structure (see
Figure 1) that influences regional volumes (13–15). Because the
PCS is less frequent in patients with psychosis (16–18) and UHR
individuals (19), spurious group differences may arise if compar-
ison groups are not matched for this morphologic variation.
In this study, we sought to address many of these limitations
by investigating ACC morphometry in a relatively large sample of
UHR individuals and healthy control subjects. We applied a
cortical surface-based protocol for parcellating the ACC (13,15),
enabling measurement of regional gray matter volume and its
constituent parameters—surface area and cortical thickness—
with submillimeter precision (20). Importantly, control subjects
and UHR-NP individuals were matched to UHR-P participants for
PCS morphology to ensure that any group differences were not
attributable to variations in cortical folding patterns. This ap-
proach enabled us to map prepsychosis ACC changes in greater
detail than previously attempted.
From the Department of Psychiatry (AF, SJW, DV, CP, MY), Melbourne Neu-
(ARY, BN), ORYGEN Research Centre, Department of Psychiatry; Depart-
ment of Psychology (LJP), School of Behavioural Science, Faculty of
PDM, MY), Department of Psychiatry, University of Melbourne; and
Howard Florey Institute (CP), The University of Melbourne, Victoria,
Centre, Levels 2 and 3, National Neuroscience Facility, 161 Barry Street,
Carlton South, Vic 3053, Australia; E-mail: email@example.com.
Received February 21, 2008; revised April 30, 2008; accepted May 29, 2008.
BIOL PSYCHIATRY 2008;64:758–765
© 2008 Society of Biological Psychiatry
Methods and Materials
UHR individuals (n ? 146) were recruited from the Per-
sonal Assessment and Crisis Evaluation (PACE) Clinic in
Melbourne, Australia. The Comprehensive Assessment of At-
Risk Mental States (CAARMS) (21), a structured clinical inter-
view designed to assess prodromal symptomatology and risk
for psychosis, was administered to all cases. Based on
CAARMS criteria, UHR individuals are characterized by one or
more of the following: 1) attenuated psychotic symptoms; 2)
brief, limited intermittent psychotic symptoms with spontane-
ous resolution; or 3) family history of psychosis accompanied
by a decline in general functioning (full criteria in ref. 22; see
also Supplement 1).
The UHR cohort was regularly monitored over a minimum
12-month period (mean ? 13; maximum ? 44 months). Partici-
pants were then divided into individuals who did (UHR-P) or did
not (UHR-NP) make the transition to psychosis using operational
criteria for determining psychosis onset (22). Final diagnoses
were made by trained psychiatrists and clinical psychologists
using the Structured Clinical Interview for DSM-IV (23) and
medical chart review. Of the 146 UHR individuals recruited, 41
became psychotic over the follow-up period (UHR-P group). Six
of the scans acquired in this group were excluded because of
image artifacts, leaving 35 UHR-P scans available for analysis.
UHR-NP individuals were then randomly selected from the
remaining 105 UHR participants and matched to each UHR-P
individual according to PCS morphology (discussed later). Sex
and age were also matched as closely as possible (diagnostic
details in Supplement 1).
Healthy control subjects came from a similar sociodemo-
graphic background to patients and were recruited by approach-
ing ancillary hospital staff and family members and through local
advertisements. The final control sample was also selected from
a larger cohort to match individually the UHR-P group for PCS
morphology, sex, and age. No participants had a history of signifi-
cant head injury, intellectual disability, steroid or substance abuse,
or neurologic disease, and control participants had no family history
of psychotic illness. Baseline symptomatology in UHR individuals
was quantified using the Scale for the Assessment of Negative
Symptoms (SANS) (24), Brief Psychiatric Ratings Scale (BPRS)
(25,26), and Hamilton Rating Scales for Depression and Anxiety
(HRSD and HRSA, respectively; Table 1) (27,28).
Scans were acquired using identical sequences on identical GE
Signa 1.5-Tesla scanners at Royal Melbourne Hospital (RMH) or
Royal Children’s Hospital (RCH), Victoria, Australia. A three-dimen-
sional volumetric spoiled gradient recoil sequence generated 124
contiguous coronal slices. Imaging parameters were time-to-echo ?
3.3 msec; time-to-repetition ? 14.3 msec; flip angle ? 30°; matrix
size ? 256 ? 256; field of view, 24 ? 24 cm; voxel dimensions,
.938 ? .938 ? 1.5 mm. The two scanners were calibrated fortnightly
of measurements. At the RMH site, 14 UHR-P, 9 UHR-NP, and 33
control individuals were scanned; the remainder were scanned at
the RCH. In studies of other brain regions in this cohort, we have
previously shown that scan site has no influence on brain volumes
matter volume: F(1,100) ? .359, p ? .550; surface area: F(1,100) ?
1.196, p ? .277; cortical thickness: F(1,100) ? .808, p ? .371].
superior rostral sulcus (SRS); bottom row presents a case with a “present” PCS and “separate” SRS. Left column presents representative sagittal slices through the
if the SRS and CS were separate, the ACCPextended from the fundus of the CS to that of the SRS. If the two sulci were continuous, the ACCPwas located on the
A. Fornito et al.
BIOL PSYCHIATRY 2008;64:758–765 759
Classification of Sulcal Variability
Before classifying PCS morphology, each participant’s im-
age was stripped of extracerebral tissue and aligned to the N27
template (31) through a 6° rigid-body transformation using
FSL software (http://www.fmrib.ox.ac.uk/fsl). The incidence
of the PCS and confluence of the superior rostral sulcus (SRS)
with the cingulate sulcus (CS) in each hemisphere were then
classified according to established criteria (13,32) to guide
region of interest (ROI) parcellation (Figure 1). UHR-P indi-
viduals were subsequently assigned to one of four categories
based on their PCS classification in each hemisphere. These
categories were as follows: present in the left and absent in the
right (n ? 14), present in both hemispheres (n ? 13), absent
in both hemispheres (n ? 3), or absent in the left and present
in the right (n ? 5). UHR-NP individuals and control subjects
were then randomly selected from a larger database and
Table 2. Mean (SD) Volume, Surface Area, and Cortical Thickness for Each Group in Each Region-of-Interest
UHR-P (n ? 35)
Volume, mean (SD) mm3
Area, mean (SD), mm2
Thickness, mean (SD), mm
UHR-NP (n ? 35)
Volume, mean (SD) mm3
Area, mean (SD), mm2
Thickness, mean (SD), mm
Control Subjects (n ? 33)
Volume, mean (SD) mm3
Area, mean (SD), mm2
Thickness, mean (SD), mm
Volume and area values are corrected for intracranial volume as detailed in Methods and Materials.
transition to psychosis onset during follow-up; UHR-P, ultra-high-risk individuals who made the transition to psychosis during follow-up. Dorsal, rostral, and
subcallosal divisions are denoted by the prefixes d-, r-, and s-, respectively.
Table 1. Demographic Details
(n ? 35)a
(n ? 35)a
(n ? 33)b
Sex, M/F (n)
Handedness, L/R/Mi (n)
Age, mean (SD), years
BPRS? total,dmean (SD)
SANS total, mean (SD)
HRSD total,emean (SD)
HRSA total,emean (SD)
Duration of symptoms before referral,
median (min.–max.), yearsf
Time between scan and psychosis
onset, median (min.–max.), years
L, left-handed; M, male; Mi, mixed-handedness; NART-IQ, National Adult Reading Test-Estimated Intelligence Quotient; R, right-handed; SANS, Scale for the
Assessment of Negative Symptoms; UHR-NP, ultra-high-risk individuals who did not make the transition to psychosis onset during follow-up; UHR-P,
ultra-high-risk individuals who made the transition to psychosis during follow-up.
aEleven UHR-P individuals and 17 UHR-NP individuals were participating in an intervention trial at the time of scanning. As such, five UHR-P and eight
UHR-NP individuals were taking low-dose risperidone (mean dose ? 1.3 mg/day) and engaged in cognitive-behavioral therapy at the time of scanning, and
antipsychotic medication at the time of scanning. The breakdown of nonantipsychotic medication in the two UHR groups was as follows: UHR-P: 10
antidepressants, 10 none, 15 data unavailable; UHR-NP: 11 antidepressants, 1 anticonvulsant, 1 anticholinergic, 1 anxiolytic, 14 none, 7 data unavailable.
bTwo UHR-P individuals could not be matched to a healthy control subject because of insufficient control women classified as having a present
paracingulate sulcus in both hemispheres (see Methods and Materials). Preliminary analyses indicated that removing the two UHR-P and two UHR-NP
cOne UHR-P individual had a NART-IQ score ? 70, but removing this participant did not change the findings, and they were retained in the final analysis.
NART-IQ data were unavailable for six UHR-P, four UHR-NP individuals, and two control subjects.
dRepresents sum of scores on BPRS scales assessing positive symptoms (i.e., scales 8, 9, 10, 11, and 15), following Ruggeri et al. Data were unavailable for
one UHR-P and one UHR-NP individual.
eData were unavailable for two UHR-P and two UHR-NP individuals.
fGroup differences assessed using the Mann-Whitney U test.
760 BIOL PSYCHIATRY 2008;64:758–765
A. Fornito et al.
individually matched to each patient on the basis of this PCS
categorization, sex, and age. Because PCS and SRS classifica-
tions tend to be related, this procedure also resulted in good
matching for SRS morphology (15).
Cortical Surface Reconstruction
The white (i.e., gray/white matter boundary) and pial (gray/
cerebrospinal fluid boundary) cortical surfaces were recon-
structed using methods described in detail by Dale, Fischl, and
colleagues (20,33), and as implemented in the Freesurfer soft-
ware package (http://surfer.nmr.mgh.harvard.edu). These sur-
face representations enabled calculation of surface area, gray
matter volume, and mean cortical thickness for each of six ACC
regions per hemisphere (discussed later) with submillimeter
precision (20) (see Figure 1). All surfaces were reconstructed
using the raw, unaligned images in native space.
Our surface-based ACC parcellation protocol has been de-
scribed in detail elsewhere (13,15). Briefly, the protocol divides
the limbic (corresponding to Brodmann area 24) and paralimbic
(primarily corresponding to Brodmann area 32) ACC (ACCLand
ACCP, respectively) into dorsal (d-ACCLand d-ACCP), rostral
(r-ACCLand r-ACCP), and subcallosal (s-ACCLand s-ACCP)
regions, yielding six ROIs per hemisphere. Boundaries separat-
ing the dorsal, rostral, and subcallosal regions were designed to
approximate previously identified subdivisions specialized for
cognitive (dorsal), affective (rostral), and autonomic/negative
emotion (subcallosal) processing (34,35). Boundaries distin-
guishing the limbic from paralimbic ACC varied in accordance
with PCS and SRS variability, following postmortem work docu-
menting how ACC cytoarchitecture changes with sulcal varia-
tions (36) (Figure 1; see ref. 13 for more details). We have
recently shown good reliabilities for this method (intraclass
correlation coefficients for all ROIs ? .8, with most ? .9) (15).
Intracranial volume (ICV) was also estimated for each individual
using a previously described method (37).
We performed four types of analysis. First, regional gray
matter volumes, surface area, and cortical thickness were
compared using separate mixed within- and between-subjects
analysis of variance (ANOVA), with hemisphere (left or right),
region (dorsal, rostral, and subcallosal), and cortex (ACCLor
ACCP) as within-subjects factors and group (UHR-P, UHR-NP,
or control) as the between-subjects factor (adding sex did not
change the findings, so we report models without it). Main
effects and interactions were evaluated with ? ? .05, using
Greenhouse-Geisser corrected degrees of freedom. Post hoc
contrasts were evaluated against a Bonferroni-adjusted alpha
to correct for multiple comparisons (i.e., the three possible
pairwise comparisons between the control, UHR-P, and
UHR-NP groups). Effect sizes (Cohen’s d) are also reported for
these differences (negative values reflect reductions in pa-
tients). In the second analysis, bivariate correlations between
gray matter changes and six clinical variables (baseline SANS,
BPRS?, HRSD, HRSA, time between scan and psychosis onset,
and duration of symptomatology before referral; see Table 1)
were evaluated against a Bonferroni-adjusted alpha (adjusted
for six tests per group) using Pearson’s r. Spearman’s rho was
used for non-Gaussian variables. The third analysis used Cox
regression to assess whether the ACC differences identified in
the ANOVA predicted time to psychosis onset in the UHR group,
independently of shared variance with baseline symptom ratings.
To prevent losing predictive information because of the conser-
vative Bonferroni corrections applied in the ANOVA, and to
avoid reliance on statistical significance as the sole inclusion
criterion, ROIs were selected for entry into the model if they
showed an effect size of at least moderate magnitude (i.e., d ?
.4). Only UHR individuals were used in this analysis because
control subjects did not complete symptom assessments. The
fourth analysis represented a preliminary investigation of subdi-
agnostic differences. The UHR-P individuals were divided into
those who developed a schizophrenia spectrum (UHR-SZ: 19
schizophrenia, 2 schizoaffective) or nonschizophreniform (UHR-
OTHER: 7 psychotic depression, 6 bipolar I, 2 psychosis not
otherwise specified) psychosis. We then ran separate ANOVAs
comparing these two groups to their respective matched control
and UHR-NP samples.
Gray matter volume and surface area estimates were cor-
rected for ICV using previously described equations (38). Cortical
thickness was not corrected for ICV because the anatomic
significance of the relationship between the two is unclear (39).
(Separate analyses that did correct for ICV yielded similar
Table 2. (continued)
A. Fornito et al.
BIOL PSYCHIATRY 2008;64:758–765 761
findings.) National Adult Reading Test—Estimated Intelligence
Quotient (NART-IQ) (40) was not a significant covariate in any of
the analyses and so was not retained in the final models.
Preonset Differences Among the UHR-P, UHR-NP,
and Control Groups
Group means for each measure are presented in Table 2.
For both gray matter volume and surface area, there was no
main effect of group [volume: F(2,100) ? .265, p ? .768; area:
F(2,100) ? 1.366, p ? .260], or any interaction between group
and region [volume: F(3.11,155.52) ? .162, p ? .927; area:
F(2.98,148.93) ? .367, p ? .775], cortex [volume: F(2,100) ?
.364, p ? .695; area: F(2,100) ? .187, p ? .830], or hemisphere
[volume: F(2,100) ? .414, p ? .662; area: F(2,100) ? .772, p ?
.465], nor was group involved in any three-way or higher-
order interactions. In a secondary analysis, we calculated area
and volume z scores, estimated relative to the control group’s
mean and standard deviation, to control for interregional
differences in size (i.e., the volume and area of dorsal and
rostral regions were much larger than those of subcallosal
areas). This analysis also found no significant group differ-
For cortical thickness, there was a significant main effect of
group [F(2,100) ? 5.276, p ? .005] and an interaction between
group, region, and cortex [F(3.05,152.66) ? 4.128, p ? .007],
but no interaction with hemisphere [F(3.19,159.37) ? 1.288,
p ? .280]. Post hoc testing revealed that, relative to healthy
control subjects, UHR-P individuals showed a significant
thickness reduction in the bilateral r-ACCP(d ? –.596 , p ?
.047, corrected). In contrast, UHR-NP individuals showed
increased thickness in the d-ACCL(d ? .592, p ? .049,
corrected), with a trend in the r-ACCL(d ? .584, p ? .054,
corrected), compared with control subjects. Comparisons be-
tween UHR-P and UHR-NP individuals revealed a bilateral
thickness reduction in the r-ACCLof the former group (d ?
–.630, p ? .029, corrected), with a trend in the s-ACCP(d ?
-.578, p ? .052, corrected). These differences are illustrated in
Figures 2 and 3.
Figure 2. (A) Effect sizes (Cohen’s d) for ultra-high-risk individuals who
relative to healthy control subjects in each subregion, collapsed across
hemispheres. Effect sizes for subgroups of UHR-P individuals who devel-
oped either a schizophrenia spectrum psychosis (UHR-SZ) or nonschizo-
phreniform (UHR-OTHER) psychosis relative to healthy control subjects are
also presented. (B) Effect sizes for the UHR-P, UHR-SZ, and UHR-OTHER
groups relative to UHR-NP individuals. ACCL? Limbic anterior cingulate
cortex; ACCP? Paralimbic anterior cingulate cortex; d-, r-, and s- denote
dorsal, rostral, and subcallosal divisions, respectively. * p ? .05, corrected.
Figure 3. Illustration of the location of anterior cingulate cortex (ACC) subregions showing relative thinning in ultra-high risk-individuals who developed
psychosis (UHR-P) compared with (A) healthy control subjects and (B) and ultra-high-risk individuals who did not develop psychosis (UHR-NP).
(C) Scatterplots illustrating the relationship between thickness in the rostral paralimbic ACC (r-ACCP) and baseline scores on the Scale for the Assessment of
ACC (r-ACCL) and baseline scores on the Hamilton Rating Scale for Anxiety (HRSA) in UHR-P (left) and UHR-NP (right) individuals.
762 BIOL PSYCHIATRY 2008;64:758–765
A. Fornito et al.
Correlations with Clinical Variables
In the UHR-P group, there was a significant correlation between
bilateral r-ACCPthickness (the region showing significant thinning
.03, corrected) that was not apparent in the UHR-NP group (r ?
.169, p ? .332, corrected; see Figure 3). The difference between
these two correlations (using Fisher’s r to z transform) was signifi-
cant (p ? .01). Conversely, a positive correlation was identified
between r-ACCLthickness (the region showing a significant differ-
ence between UHR-P and UHR-NP individuals) and HRSA ratings
in UHR-NP individuals (r ? .54, p ? .01, corrected), but not
UHR-P participants (r ? –.075, p ? .679, corrected). The
difference between these two correlations was also significant
(p ? .01). No other significant correlation between thickness and
symptom or clinical measures was identified.
Prediction of Psychosis Onset During Follow-Up
Four regions showed an effect size ? .4 in contrasts between
UHR-P and UHR-NP individuals: the r-ACCL, r-ACCP, s-ACCL, and
s-ACCP. Thickness values were averaged across these regions
and entered into a Cox regression as predictors, along with
baseline SANS and HRSA scores (i.e., the only two symptom
scales showing any correlations with ACC measures). This anal-
ysis revealed that thickness in these regions was a significant
predictor of time to psychosis onset (Wald ? 5.149, ? ? .196,
95% confidence interval [CI] ? .05–.80, p ? .023), suggesting that,
for every 1-mm decrease in thickness, there was an approximately
20% increase in risk for psychosis (or, conversely, a ?20% decrease
in risk for psychosis for every 1-mm thickness increase, given that
the relationship is driven both by thinning in UHR-P individuals and
thickening in UHR-NP individuals; see Discussion). The mean
thickness difference between the two groups in these regions was
approximately .2 mm, suggesting the observed thickness changes
related to an average risk increase of approximately 4%. The SANS
ratings also predicted time to transition (Wald ? 5.485, ? ? 1.027,
95% CI ? 1.00–.1.05, p ? .019), but HRSA scores did not (Wald ?
.304, ? ? 1.011, 95% CI ? .97–1.05, p ? .581). The thickness
measures remained significant as predictors of outcome (? ? .218,
95% CI ? .05–.99, p ? .048) after the other symptom measures
(HRSD and BPRS?) were added to the model.
The ANVOVA comparing UHR-SZ individuals to their matched
control subjects and UHR-NP groups revealed a three-way interac-
tion among diagnosis, region, and cortex [F(3.14,94.05) ? 3.639,
p ? .014]. Post hoc contrasts indicated that, compared with control
subjects, the UHR-SZ group showed reduced thickness in the
r-ACCP(d ? –.790, p ? .046, corrected) and s-ACCL(d ? –.782,
p ? .049, corrected). Relative to UHR-NP individuals, the UHR-SZ
group showed reduced r-ACCL(d ? –.824, p ? .035, corrected)
and s-ACCP(d ? –.906, p ? .017, corrected) thickness. No
significant differences were identified between UHR-OTHER
individuals and their matched control or UHR-NP groups. Direct
comparison between UHR-SZ and UHR-OTHER individuals in a
subset that could be matched for PCS morphology (n ? 13)
revealed no differences.
Identifying the neurobiological changes that precede psycho-
sis onset represents an important step in understanding the
pathogenesis of psychotic disorders. In this study, we have
demonstrated that anatomic changes in the ACC 1) are apparent
before the onset of frank psychosis, 2) distinguish between UHR
individuals who subsequently do and do not develop a psychotic
episode, and 3) are relatively specific to individuals who develop
a schizophrenia spectrum disorder. Consistent with previous
findings (7,15), our data also indicate that cortical thickness
represents a particularly sensitive metric for detecting anatomic
changes, possibly because it shows less variability across indi-
viduals than area and volume (15). Importantly, our procedure
for matching groups for local sulcal and gyral variability ensured
these changes are unlikely to reflect differences in cortical
folding patterns. Together these findings indicate that ACC
abnormalities represent a risk marker of transition to psychosis.
ACC Abnormalities Preceding Psychosis Onset
Comparison of UHR-P individuals with healthy control sub-
jects revealed that thinning of the r-ACCPprecedes the onset of
frank psychosis. The localization of these changes is interesting
in light of our recent work demonstrating bilateral thinning of the
paralimbic region in first episode schizophrenia (8), and longi-
tudinal research in childhood-onset schizophrenia suggesting the
earliest postonset changes appear in paralimbic regions and
spread to engulf the limbic ACC over a 5-year period (41). We
have also found evidence for longitudinal gray matter reductions
in dorsal paralimbic regions during the transition to psychosis
(9). Together these findings suggest a timetable for the progres-
sion of ACC abnormalities in schizophrenia, whereby the earliest
reductions emerge in the rostral paralimbic region during the
prodrome, extend across the dorsal and subcallosal paralimbic
regions following psychosis onset, and spread to encompass
limbic regions with continued illness duration.
Evidence that the r-ACCPplays an important role in social
cognitive processing, particularly mentalizing others’ intentions
(42), suggests abnormalities in the area lead to difficulties
interacting with others. This may underlie the increasing social
withdrawal that is characteristic of the psychosis prodrome
(43,44). Recent data suggest genetic high-risk individuals display-
ing subthreshold psychotic symptoms show poorer performance
on theory-of-mind tasks than those without such symptoms (45).
In this context, our finding of an inverse correlation between
r-ACCPthickness and negative symptoms in the UHR-P group
suggests the functional consequences of these anatomic changes
include a diminished engagement with the external world.
However, determining whether the r-ACCpchanges cause eleva-
tions in negative symptoms or whether both are manifestations
of some other trait marker requires further longitudinal investi-
gation. Given that we found no correlation between r-ACCP
thickness and length of symptoms before referral, the thickness
changes appear related to the magnitude rather than duration of
ACC Changes Predictive of the Transition to Psychosis
Comparison between UHR-P and UHR-NP individuals indi-
cated that changes in the r-ACCLand, to a lesser extent, the
rostral paralimbic and limbic and paralimbic subcallosal regions,
differentiated between the two groups. Cox regression indicated
these thickness differences predicted time to psychosis onset
independently of any correlations with baseline symptom rat-
ings. Examination of these differences relative to healthy control
subjects indicated the association was driven by abnormal thick-
ening in the UHR-NP group, in addition to thinning in the UHR-P
group, suggesting that baseline differences between high-risk
individuals who do and do not subsequently become psychotic
may result from abnormalities in either group, not just prepsy-
chotic individuals. Our previous study of UHR individuals (9) did
A. Fornito et al.
BIOL PSYCHIATRY 2008;64:758–765 763
not examine control subjects, and although Borgwardt et al. (10)
and Job et al. (12) both employed control groups in their studies,
the former only compared them to the high-risk cohort as a
whole (irrespective of diagnostic outcome), with the latter com-
paring longitudinal, but not baseline, changes relative to control
subjects (although an earlier study examined baseline differences
between control subjects and the high-risk group as a whole)
(11). Our current findings therefore caution against attributing
preonset differences between high-risk individuals who do and
do not subsequently develop psychosis to abnormalities in the
former group unless healthy control subjects are also examined.
An intriguing possibility arising from these data is that abnormal
thickening in the UHR-NP group confers increased resilience
against transition to psychosis, although the positive correlation
observed between thickness and anxiety levels suggests that it
predisposes to other psychopathology.
The differences between the UHR-P and UHR-NP groups are
unlikely to be due to variations in their baseline clinical charac-
teristics, given that ACC measures predicted time to psychosis
onset independently of any shared variance with symptom
measures. The two groups were well-matched demographically,
and while a small proportion of UHR individuals were also
enrolled in an intervention trial, the relative proportions in each
group receiving each intervention type were similar (Table 1).
Moreover, the dissociation observed with respect to correlations
between regional ACC abnormalities and symptom measures in
the two groups suggests that the anatomic changes are related to
distinct pathophysiologic processes related to their divergent
The changes observed in the UHR-P group were primarily
driven by individuals who went on to develop a schizophrenia
spectrum psychosis. Effect sizes relative to both the control and
UHR-NP groups were larger when analyses were restricted to the
UHR-SZ group, and the differences were generally in the same
subregions. In contrast, effect sizes for contrasts involving the
UHR-OTHER group were relatively small.
The only other study to investigate preonset brain changes
related specifically to risk for schizophrenia found no evidence for
baseline differences between those who eventually developed the
illness and those who did not (12). One reason for this discrepancy
is that thickness, but not volume, may represent a specific marker of
risk for psychosis, consistent with reports that other, nonvolumetric
brain changes predict transition (46). Further reasons for our
discrepant findings include our attempt to account for the con-
founding influence of PCS variability and the possibility that genetic
high-risk individuals may show different neurobiological changes
from those at clinical high risk. Although our multifaceted approach
to defining risk is likely to be more representative of the spectrum
of psychosis patients (i.e., not all such individuals possess a family
history), it is possible that different risk categories are associated
us to stratify analyses by risk category.
Images were acquired on two scanners, although it is unlikely
that this influenced our findings for several reasons. First, the
scanners were of an identical make and calibrated fortnightly.
Second, we have previously shown that differences between
these two scanners have no effect on a variety of neuroanatomic
measures (29,30), including those used in this study. Third, if the
differences were due to scan site effects, it is unclear why UHR-P
and UHR-NP should display a distinct and regionally specific
profile of anatomic changes relative to healthy control subjects,
why these should be correlated with specific symptom dimen-
sions, and why they should occur for thickness but not volume or
surface area. Finally, the cortical surface reconstruction tech-
niques and resulting measures we used are robust to variations
across scanners and acquisition protocols (20,47).
We individually matched groups for PCS morphology to
control for the confounding influence of ACC anatomic variabil-
ity (13,15). This matching procedure introduced no sampling
biases in UHR-P group, which comprised all UHR-P individuals
for whom MRI data of sufficient quality were available. Although
healthy control subjects and UHR-NP individuals were selected
from a larger sample to match the UHR-P individuals for PCS
morphology, there were no significant differences between those
who were and were not selected in terms of age or NART-IQ. In
any case, the matching procedure would have minimized group
differences, because they were chosen on the basis of possessing
similar brain morphology. However, individuals do not present
in clinical settings after having been selected for some neuroana-
tomic characteristic. Thus although our findings indicate that ACC
changes are able to discriminate between UHR individuals who do
and do not subsequently develop psychosis independently of
differences in baseline symptom profiles or variations in cortical
folding patterns, their practical utility in predicting which high-risk
individuals will make the transition to psychosis remains a topic for
further investigation in unselected samples. We also note that
although some UHR-NP individuals may have developed psychosis
after our follow-up period, the majority of transitions occur within
the first year (48). We are currently conducting a longer-term
follow-up of this cohort to assess their outcomes.
In summary, we have demonstrated that anatomic abnormal-
ities of the ACC are present before psychosis onset and are
relatively specific to risk for schizophrenia spectrum disorders.
Further work that examines the longitudinal progression of these
changes and integrates them with changes observed in other
brain regions (10,12,49) will increase the predictive power of
neuroanatomic risk markers and elucidate underlying patho-
Neuroimaging analysis was facilitated by the Neuropsychia-
try Imaging Laboratory managed by Ms. Bridget Soulsby at the
Melbourne Neuropsychiatry Centre and supported by Neuro-
sciences Victoria. The research was supported by the Melbourne
Neuropsychiatry Centre (Sunshine Hospital), Department of Psy-
chiatry, the University of Melbourne, the National Health and
Medical Research Council (ID 236175; 350241), and the Ian
Potter Foundation. AF was supported by a JN Peters Fellowship
and a National Health and Medical Research Council (NHMRC)
CJ Martin Fellowship (ID: 454797). SJW was supported by a
NHMRC Clinical Career Development Award and a National
Alliance for Research on Schizophrenia and Depression Young
Investigator Award. MY was supported by a NHMRC Clinical
Career Development Award (ID: 509345).
The authors reported no biomedical financial interests or
potential conflicts of interest.
Supplementary material cited in this article is available
of Cingulate Cortex and Limbic Thalamus: A Comprehensive Handbook.
764 BIOL PSYCHIATRY 2008;64:758–765
A. Fornito et al.
2. Yücel M, Wood SJ, Fornito A, Riffkin J, Velakoulis D, Pantelis C (2003): Download full-text
Anterior cingulate dysfunction: Implications for psychiatric disorders?
3. Sanders GS, Gallup GG, Heinsen H, Hof PR, Schmitz C (2002): Cognitive
deficits, schizophrenia, and the anterior cingulate cortex. Trends Cogn
4. Goldstein JM, Goodman JM, Seidman LJ, Kennedy DN, Makris N, Lee H,
tural magnetic resonance imaging. Arch Gen Psychiatry 56:537–547.
5. Benes FM, Vincent SL, Todtenkopf M (2001): The density of pyramidal
and nonpyramidal neurons in anterior cingulate cortex of schizo-
phrenic and bipolar subjects. Biol Psychiatry 50:395–406.
7. Fornito A, Malhi GS, Lagopoulos J, Ivanovski B, Wood SJ, Saling MM, et al.
8. Fornito A, Yucel M, Wood SJ, Adamson C, Velakoulis D, Saling MM,
et al. (2008): Surface-based morphometry of the anterior cingulate cor-
tex in first episode schizophrenia. Hum Brain Mapp 29:478–489.
9. Pantelis C, Velakoulis D, McGorry PD, Wood SJ, Suckling J, Phillips LJ, et
al. (2003): Neuroanatomical abnormalities before and after onset of
psychosis: A cross-sectional and longitudinal MRI comparison. Lancet
10. Borgwardt SJ, Riecher-Rossler A, Dazzan P, Chitnis X, Aston J, Drewe M,
et al. (2007): Regional gray matter volume abnormalities in the at risk
mental state. Biol Psychiatry 61:1148–1156.
11. Job DE, Whalley HC, McConnell S, Glabus M, Johnstone EC, Lawrie SM
(2003): Voxel-based morphometry of grey matter densities in subjects
at high risk of schizophrenia. Schizophr Res 64:1–13.
12. Job DE, Whalley HC, Johnstone EC, Lawrie SM (2005): Grey matter
13. Fornito A, Whittle S, Wood SJ, Velakoulis D, Pantelis C, Yucel M (2006):
The influence of sulcal variability on morphometry of the human ante-
rior cingulate and paracingulate cortex. Neuroimage 33:843–854.
14. Paus T, Otkay N, Caramanos Z, MacDonald D, Zijdenbos A, D’Avirro D,
et al. (1996): In vivo morphometry of the intrasulcal gray matter in the
asymmetries, gender differences and probability maps. J Comp Neurol
15. Fornito A, Wood SJ, Whittle S, Fuller J, Adamson C, Saling MM, et al.
(2008): Variability of the paracingulate sulcus and morphometry of the
medial frontal cortex: Associations with cortical thickness, surface area,
volume, and sulcal depth. Hum Brain Mapp 29:222–236.
Paracingulate morphologic differences in males with established
schizophrenia: A magnetic resonance imaging morphopmetric study.
17. Le Provost JB, Bartres-Faz D, Paillere-Martinot ML, Artiges E, Pappata S,
Recasens C, et al. (2003): Paracingulate sulcus morphology in men with
18. Fornito A, Malhi GS, Lagopoulos J, Ivanovski B, Wood SJ, Velakoulis D,
the anterior cingulate cortex in bipolar disorder. Acta Psychiatr Scand
19. Yücel M, Wood SJ, Phillips LJ, Stuart GW, Smith DJ, Yung A, et al. (2003):
risk of developing a psychotic illness. Brit J Psychiatry 182:518–524.
cortex from magnetic resonance images. Proc Natl Acad Sci U S A 97:
21. Yung AR, Yuen HP, McGorry PD, Phillips LJ, Kelly D, Dell’Olio M, et al.
(2005): Mapping the onset of psychosis: The Comprehensive Assess-
22. Yung A, Phillips L, McGorry PD (2004): Treating Schizophrenia in the
Prodromal Phase. London: Taylor & Francis.
23. First MB, Spitzer RL, Gibbon M, Williams JB (1998): Structured Clinical
Interview for DSM-IV. Washington, DC: American Psychiatric Press.
24. Andreasen NC (1982): Negative symptoms in schizophrenia. Definition
26. McGorry PD, Goodwin RJ, Stuart GW (1988): The development, use, and
reliability of the brief psychiatric rating scale (nursing modification)—an
assessment procedure for the nursing team in clinical and research set-
27. Hamilton M (1959): The assessment of anxiety states by rating. Br J
28. Hamilton M (1960): A rating scale for depression. J Neurol Neurosurg
29. Velakoulis D, Wood SJ, Wong MT, McGorry PD, Yung A, Phillips L, et al.
(2006): Hippocampal and amygdala volumes according to psychosis
stage and diagnosis: A magnetic resonance imaging study of chronic
schizophrenia, first-episode psychosis, and ultra-high-risk individuals.
30. Phillips LJ, Velakoulis D, Pantelis C, Wood S, Yuen HP, Yung AR, et al.
risk of psychosis. Schizophr Res 58:145–58.
31. Holmes CJ, Hoge R, Collins L, Woods R, Toga AW, Evans AC (1998):
Enhancement of MR images using registration for signal averaging.
32. Yücel M, Stuart GW, Maruff P, Velakoulis D, Crowe SF, Savage G, et al.
(2001): Hemispheric and gender-related differences in the gross mor-
teers: An MRI morphometric study. Cereb Cortex 11:17–25.
33. Dale AM, Fischl B, Sereno MI (1999): Cortical surface-based analysis I.
Segmentation and surface reconstruction. Neuroimage 9:179–194.
34. Bush G, Luu P, Posner MI (2000): Cognitive and emotional influences in
anterior cingulate cortex. Trends Cogn Sci 4:215–222.
late cortex to behaviour. Brain 118:279–306.
36. Vogt BA, Nimchinsky EA, Vogt LJ, Hof PR (1995): Human cingulate cor-
tex: Surface features, flat maps, and cytoarchitecture. J Comp Neurol
An optimized method for estimating intracranial volume from mag-
netic resonance images. Magn Reson Med 44:973–977.
38. Free SL, Bergin PS, Fish DR, Cook MJ, Shorvon SD, Stevens JM (1995):
Methods for normalization of hippocampal volumes measured with
40. Nelson HE, O’Connell A (1978): Dementia: The estimation of premorbid
41. Vidal CN, Rapoport JL, Hayashi KM, Geaga JA, Sui Y, McLemore LE, et al.
adolescents with schizophrenia. Arch Gen Psychiatry 63:25–34.
42. Amodio DM, Frith CD (2006): Meeting of minds: The medial frontal
cortex and social cognition. Nat Rev Neurosci 7:268–277.
44. Jackson HJ, McGorry PD, Dudgeon P (1995): Prodromal symptoms of
schizophrenia in first-episode psychosis: Prevalence and specificity.
EC, Lawrie S (2006): A neuropsychological investigation into “theory of
mind” and enhanced risk of schizophrenia. Psychiatry Res 144:29–37.
et al. (2007): Increased prefrontal gyrification in a large high-risk cohort
characterizes those who develop schizophrenia and reflects abnormal
prefrontal development. Biol Psychiatry 62:722–729.
47. Han X, Jovicich J, Salat D, van der Kouwe A, Quinn B, Czanner S, et al.
tical thickness: The effects of field strength, scanner upgrade and man-
ufacturer. Neuroimage 32:180–194.
48. Yung AR, Phillips LJ, Yuen HP, McGorry PD (2004): Risk factors for psy-
chosis in an ultra high-risk group: Psychopathology and clinical fea-
tures. Schizophr Res 67:131–142.
(2005): Pituitary volume predicts future transition to psychosis in indi-
A. Fornito et al.
BIOL PSYCHIATRY 2008;64:758–765 765