Cortical and Subcortical Brain Morphometry Abnormalities in Youth at Clinical High-Risk for Psychosis and Individuals with Early Illness Schizophrenia

Abstract

Neuroimaging studies have documented morphometric brain abnormalities in schizophrenia, but less is known about them in individuals at clinical high-risk for psychosis (CHR-P), including how they compare with those observed in early schizophrenia (ESZ). Accordingly, we implemented multivariate profile analysis of regional morphometric profiles in CHR-P (n=89), ESZ (n=93) and healthy controls (HC; n=122). ESZ profiles differed from HC and CHR-P profiles, including 1) cortical thickness: significant level reduction and regional non-parallelism reflecting widespread thinning, except for entorhinal and pericalcarine cortex, 2) basal ganglia volume: significant level increase and regional non-parallelism reflecting larger caudate and pallidum, and 3) ventricular volume: significant level increase with parallel regional profiles. CHR-P and ESZ cerebellar profiles showed significant non-parallelism with HC profiles. Regional profiles did not significantly differ between groups for cortical surface area or subcortical volume. Compared to CHR-P followed for ≥18 months without psychosis conversion (n=31), CHR-P converters (n=17) showed significant non-parallel ventricular volume expansion reflecting specific enlargement of lateral and inferolateral regions. Antipsychotic dosage in ESZ was significantly correlated with frontal cortical thinning. Results suggest that morphometric abnormalities in ESZ are not present in CHR-P, except for ventricular enlargement, which was evident in CHR-P who developed psychosis.

1. Introduction

Many prior magnetic resonance imaging (MRI) studies have documented morphometric brain abnormalities in schizophrenia (e.g., Haijma et al., 2013; Harvey et al., 1993; Kempton et al., 2010; Koshiyama et al., 2022; McCarley et al., 1999; Shenton et al., 2001; van Erp et al., 2016; van Erp et al., 2018; Wright et al., 2000), with abnormalities becoming more widespread over the illness course (Jung et al., 2011; van Haren et al., 2011). However, there is much research documenting brain morphometric alterations in schizophrenia across the illness course, from first-episode to chronic patients, less is known about these morphometric measures in individuals at clinical high-risk for psychosis (CHR-P) and how they compare with measures from schizophrenia patients early in their illness course (ESZ). Although still actively debated (Borgwardt et al., 2009; DeLisi, 2022; Fatemi & Folsom, 2009; Fusar-Poli et al., 2013; Mathalon et al., 2003; van Haren et al., 2011; Zipursky et al., 2013), many longitudinal studies have shown progressive worsening of morphometric brain abnormalities over time in schizophrenia over and above the effects of normal aging, consistent with progressive pathophysiological processes (Andreasen et al., 2011; Cahn et al., 2002; DeLisi, 2008; Dietsche et al., 2017; Fusar-Poli et al., 2013; Ho et al., 2003; Kempton et al., 2010; Mathalon et al., 2001; Vita et al., 2012). Moreover, while the presence of morphometric brain abnormalities in first-episode schizophrenia supports the prevailing theory they largely arise from faulty neurodevelopment (Andreasen et al., 2011; Cahn et al., 2002; Ho et al., 2003; Smucny et al., 2022; Weinberger, 1995), their neurodevelopmental origins are not incompatible with a progressive pathophysiology worsening these abnormalities over the illness course (Mathalon et al., 2003; Rapoport et al., 2005).

Previous neuroimaging studies have specifically documented cortical abnormalities early in the course of schizophrenia. Many studies have documented cortical thinning in ESZ, especially in frontal and temporal regions (e.g., Akudjedu et al., 2020; Crespo-Facorro et al., 2011; Gallardo-Ruiz et al., 2019; Hua & Mathalon, 2022; Lin et al., 2019; Rapoport et al., 2005; Wen et al., 2021; Zhao et al., 2022). In contrast, only three studies have examined regional cortical surface area, with two finding significant differences (Haring et al., 2016; Hua & Mathalon, 2022) and one finding no differences (Crespo-Facorro et al., 2011). Multiple studies have also implicated subcortical/ventricular abnormalities in ESZ, including reports of decreased caudate, hippocampus, amygdala, putamen, and thalamus volumes (Crespo-Facorro et al., 2009; Ebdrup et al., 2010; Fan et al., 2019; Hua & Mathalon, 2022) and enlarged 3^rd and lateral ventricular volumes (HC, Cahn et al., 2002; Crespo-Facorro et al., 2009; Hua & Mathalon, 2022; Rosa et al., 2010). With regard to the cerebellum, there is mixed evidence of decreased cerebellar volume in ESZ with some finding significant differences (Bottmer et al., 2005; Chan et al., 2011; Kim et al., 2018) and others not (Morimoto et al., 2021; Steen et al., 2006). However, many studies focused on selective a priori regions, rather than examining deficits across a more comprehensive regional profile, making it difficult to evaluate inconsistent volumetric findings across studies.

A potential confound in most morphometric brain imaging studies of schizophrenia, including ESZ, is that most patients are currently taking, and have a history of treatment with, antipsychotic medications. Both animal and human studies provide evidence that dopamine D2 antagonists can induce brain morphometric changes (e.g., Dorph-Petersen et al., 2005; Konopaske et al., 2008; Lieberman et al., 2005; Roiz-Santiañez et al., 2015), including increased caudate volume (Albacete et al., 2019; Andersen et al., 2020; Chakos et al., 1994; Lang et al., 2004; Roiz-Santiañez et al., 2015) and, less consistently, reduced frontal cortex thickness (Cannon et al., 2016; Haijma et al., 2013; Lesh et al., 2015; Pies, 2018; Roiz-Santiañez et al., 2015). Nonetheless, cortical thickness reduction has been found in both antipsychotic-naïve and medicated ESZ patients, suggesting this reduction results, at least partly, from primary disease pathophysiology rather than exclusively resulting from antipsychotic exposure (Andreasen et al., 2011).

MRI studies examining whether morphometric brain abnormalities are present prior to psychosis onset in CHR-P individuals, as predicted by prevailing neurodevelopmental models of schizophrenia, have yielded mixed findings (e.g., Bois, Ronan, et al., 2015; Brent et al., 2013; Cannon et al., 2015; Jalbrzikowski et al., 2021). The strongest evidence comes from studies of cortical thickness (Bois, Whalley, et al., 2015; Brent et al., 2013; Collins et al., 2022; Jalbrzikowski et al., 2021; Zhao et al., 2022). Regional cortical thinning has been found in CHR-P individuals at baseline compared to HC, particularly with regard to bilateral medial frontal cortex (meta-analysis, Zhao et al., 2022). However, there is much variability across studies with some studies finding differences (Chung et al., 2019; Collins et al., 2022; del Re et al., 2021; Gisselgård et al., 2018; Jalbrzikowski et al., 2021; Jung et al., 2011; Kwak et al., 2019; Tomyshev et al., 2019), but not others (Bakker et al., 2016; Bois, Ronan, et al., 2015; Buechler et al., 2020; Cannon et al., 2015; Klauser et al., 2015; Ziermans et al., 2012). Relatively few CHR-P studies have examined cortical surface area and/or subcortical/ventricular/cerebellar volumes, which have great relevance to brain organization and functioning and may show important differential patterns of abnormalities. Recent evidence has been accumulating showing none to few cortical surface area abnormalities in CHR-P ((del Re et al., 2021; Jalbrzikowski et al., 2021; Tomyshev et al., 2019); but see (Bois, Ronan, et al., 2015)). There is mixed evidence of significant subcortical and/or ventricular volume abnormalities in CHR-P (Berger et al., 2017; Cannon et al., 2015; Chung et al., 2019; Jalbrzikowski et al., 2021; Konishi et al., 2018; Sasabayashi et al., 2020). Further, there is also mixed evidence of cerebellar volume abnormalities in CHR-P (Chan et al., 2011; Morimoto et al., 2021; Roman-Urrestarazu et al., 2014).

In addition, the potential for morphometric brain measures to serve as predictive biomarkers of future psychosis risk in CHR-P individuals remains unsettled, although this possibility may explain some of the inconsistent findings associated with CHR-P comparisons with HC. For example, morphometric abnormalities present during the CHR-P period may be subtle initially and only worsen over time. Alternatively, weak or inconsistent morphometric abnormalities in CHR-P individuals as a group may, in fact, reflect more prominent abnormalities present only in the small subset of CHR-P individuals who subsequently convert to a psychotic disorder (e.g., Andreou & Borgwardt, 2020; Collins et al., 2022; Ziermans et al., 2012). Consistent with the latter, decreased baseline cortical thickness (Buechler et al., 2020; del Re et al., 2021; Jalbrzikowski et al., 2021; Merritt et al., 2021; Walterfang et al., 2008) and steeper rates of cortical thinning (Cannon et al., 2015; Collins et al., 2022) have been found in CHR-P individuals who subsequently converted to psychosis relative to CHR-P non-converters. Additionally, baseline enlargement of 3^rd and lateral ventricle volumes has been found in CHR-P converters relative to non-converters. (Cannon et al., 2015; Chung et al., 2017; Sasabayashi et al., 2020) Thus, structural brain abnormalities tend to be evident in those who subsequently convert to a psychotic disorder; however, this impression comes from a relatively small set of studies, many of which were underpowered or examined only a few select brain regions.

Accordingly, in the current structural MRI study, we examined multiple cortical and subcortical regional morphometric measures in CHR-P, ESZ, and HC individuals. Using multivariate profile analyses, we examined whether regional profiles of morphometric measures showed significant “non-parallelism” (i.e., pattern differences) or overall elevation or “level” differences between groups, with separate analyses conducted for cortical thickness, cortical surface area, subcortical volumes, and ventricular volumes. Moreover, longitudinal clinical follow-up data on CHR-P individuals allowed us to compare morphometric regional profiles in those who subsequently converted to psychosis relative to those who did not but were followed clinically for at least 18 months.

2. Methods

2.1. Participants

CHR-P individuals (n=89) were recruited from early psychosis research clinics at the University of California, San Francisco (UCSF; n=62) and the University of California, Davis (UCD; n=27). CHR-P individuals met Criteria of Psychosis-Risk Syndromes (COPS), based on the Structured Interview for Psychosis-Risk Syndromes (SIPS, McGlashan et al., 2010; Miller et al., 2003; Miller et al., 2002), for one or more of three non-mutually exclusive sub-syndromes: attenuated psychotic symptoms, brief intermittent psychotic states, and genetic risk with deterioration in social/occupational functioning. Within the CHR-P group, 17 individuals converted to a psychotic disorder and 31 did not convert to a psychotic disorder but had been followed clinically for at least 18 months from their scan date. The clinical outcomes of the 41 CHR-P who dropped out of the study prior to their 18-month follow-up assessment could not be determined with confidence, so they were excluded from the CHR-P clinical outcome analyses (but were retained in baseline group comparisons of CHR-P with ESZ and HC). Current symptom severity (positive, negative, disorganized, general) was assessed using the Scale of Psychosis-Risk Symptoms (SOPS, McGlashan et al., 2010; Miller et al., 2003; Miller et al., 2002). A subset of CHR-P individuals was taking antipsychotic medication at study entry (n=13/89; 14.61%) but were included because their past symptoms had never reached the threshold for full psychosis.

ESZ patients (n=93) were recruited from early psychosis clinics at UCSF (n=78) and UCD (n=15) and met criteria for a current diagnosis of schizophrenia or schizoaffective disorder based on the Structured Clinical Interview for DSM-IV (SCID, First et al., 2002), with illness onset within the last five years. The ESZ group had a mean illness duration of 2.35 years (SD = 1.45 years), with 48% being within the first 2 years of illness onset and 73% within the first 3 years of illness onset. Current symptom severity (positive, negative, general) was assessed with the Positive and Negative Syndrome Scale (PANSS, Kay et al., 1987). Most ESZ patients were prescribed antipsychotic medication (n=75/93; 80.65%). HC participants (n=122) were recruited from the local UCSF (n=90) and UCD (n=32) communities. HC did not meet criteria for any Axis I diagnosis based on the SCID (for HC >16 years old) or the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children, Present and Lifetime Version (for HC ≤16 years old) (Kaufman et al., 1997).

Across both sites, participants had to be in good physical health, fluent in English, and between 12–35 years old. Exclusion criteria included a past year DSM-IV substance dependence diagnosis (except nicotine), head injury with loss of consciousness, neurological illness, or a first-degree relative with a psychotic illness (for HC). Study protocols were in accordance with UCSF’s and UCD’s Institutional Review Boards and the Declaration of Helsinki. Participants provided written informed consent, or in the case of minors, parental written consent and participant assent.

2.2. Structural MRI acquisition and processing

Participants were scanned at neuroimaging centers at UCSF and UCD using identical Siemens 3T TIM TRIO scanners and high-resolution structural T1-weighted MPRAGE image acquisition sequences using the following parameters: repetition time (TR)=2,300 ms, echo time (TE)=2.95 ms, flip angle=9°, field of view (FOV)=256×256, slice thickness=1.20 mm.

MRI images were processed using FreeSurfer (version 5.1, Han et al., 2006; Jovicich et al., 2006, see Supplemental Methods). Consistent with multi-site ENIGMA studies arguing for analytic standardization (van Erp et al., 2016; van Erp et al., 2018), the same FreeSurfer-provided parcellations were used to define regions of interest (ROIs). Cortical ROIs were based on the Desikan-Killiany atlas (Desikan et al., 2006), and subcortical/ventricular ROIs were from FreeSurfer’s atlas (Fischl et al., 2002). Thickness and surface area of cortical regions (68 unilateral ROIs–34 in each hemisphere) and volume of subcortical (14 unilateral ROIs–7 in each hemisphere), ventricular (5 unilateral ROIs–2 in each hemisphere, plus 3rd ventricle), and cerebellar (4 unilateral ROIs–2 in each hemisphere) regions were calculated.

2.3. Statistical analyses

2.3.1. Z-scoring adjustment procedure

To account for normal brain maturation (i.e., age, Mills et al., 2016), sex (Barnes et al., 2010; Hogstrom et al., 2013; Luders et al., 2004; Ritchie et al., 2018), scanning site, and intracranial volume (ICV, Barnes et al., 2010; Im et al., 2008; Toro et al., 2008), the effects of these variables were statistically removed from the morphometric brain measures. As cortical thickness is not as strongly correlated with ICV as surface area and volume, thickness measures were not statistically adjusted for ICV (Barnes et al., 2010) (FreeSurfer guidelines, https://surfer.nmr.mgh.harvard.edu/fswiki/eTIV). For the remaining regional morphometric measures, the effects of age, sex, ICV, and site were modeled in the HC group, and resulting parameter estimates were used to derive predicted values for each participant, reflecting the value expected for a healthy individual of a specific biological sex, age, and ICV assessed at a specific study site. By subtracting observed values from these predicted values and then dividing by the standard error of regression from the HC regression model (i.e., standard deviation of HC data points from the HC regression line), morphometric scores were transformed to adjusted z-scores (see Supplemental Methods, Hua et al., 2022; Pfefferbaum et al., 1992; Roach et al., 2020). Adjusted z-scores represent the participant’s deviation, in standard units, from the value for a HC of the same age, sex, ICV (except for cortical thickness measures), and study site as the participant.

2.3.2. Structural morphometric analyses

Using multi-factorial repeated measures multivariate analyses of variance (rmMANOVA), group differences for each type of morphometric ROI profile were assessed, including 1) cortical thicknesses, 2) cortical surface areas, 3) basal ganglia volumes (caudate, putamen, pallidum), 4) subcortical volumes (accumbens, hippocampus, amygdala, thalamus) 5) ventricular volumes, and 6) cerebellar volumes. Groups were compared on two aspects of each ROI profile: 1) non-parallelism of the profiles (i.e., Group x Region interaction effect) and 2) level differences between the profiles (i.e., Group main effect). Basal ganglia ROIs, which have been shown to be enlarged in schizophrenia due to the effect of antipsychotic medications (Albacete et al., 2019; Andersen et al., 2020; Chakos et al., 1994; Lang et al., 2004), were analyzed separately from other subcortical ROIs that typically show reductions in schizophrenia (van Erp et al., 2016). Given that adjusted z-scores re-express structural morphometric values as deviations from the expected normative value, profiles of z-scores for the HC group are flat and the mean for each ROI is approximately zero, whereas the profiles of CHR-P and ESZ groups reflect the extent and regional pattern of variation from the age norms derived from the HC group.

Initially, for each type of morphometric measure (cortical thickness, cortical surface area, basal ganglia volume, subcortical volume, ventricular volume, cerebellar volume), 3-way Group (HC, CHR-P, ESZ) x Hemisphere (Left, Right) x ROI rmMANOVA models were run. Due to non-significant 3-way interactions for all six models (ps>.119), ROIs were averaged between hemispheres, resulting in 34 cortical, 3 basal ganglia, 4 subcortical, 3 ventricular, and 2 cerebellar bilateral ROIs. Subsequent 2-way Group x ROI models focused on tests of regional profile non-parallelism between groups (i.e., Group x ROI interaction) and tests of profile level or elevation differences between groups (i.e., Group main effect). Significant non-parallelism of profiles was followed-up with 2-group Group x ROI rmMANOVAs comparing the profiles of HC vs. CHR-P, HC vs. ESZ, and CHR-P vs. ESZ. Lastly, following-up on significant non-parallelism of the 2-group rmMANOVAs, we examined Bonferroni-corrected simple effects comparing groups on each bilateral ROI.

2.3.3. Prediction of clinical outcomes in clinical high risk for psychosis group

We divided the CHR-P group into those who converted to a psychotic disorder and those who did not convert to a psychotic disorder but had been assessed clinically for at least 18 of the 24 months comprising the study follow-up period. Similar to the 3-group rmMANOVA models, CHR-P converter vs. non-converter morphometric regional profiles were compared using Group x ROI multivariate profile analyses

2.3.4. Antipsychotic medication and symptom severity analyses

To examine relationships between morphometric measures and clinical variables such as symptom severity ratings or antipsychotic medication without having to adjust for an excessive number of correlation tests, we first reduced the ROI measures to a smaller set of regional components by performing principal component analyses with promax rotation on unilateral ROIs separately for cortical thickness, cortical surface area, basal ganglia volume, subcortical volume, ventricular volume, and cerebellar volume in the combined sample of CHR-P, ESZ, and HC (for full details, see Supplemental Methods, Supplemental Tables 1–5, Supplemental Figures 1 and 2. Associations of these extracted ROI components with symptom severity and antipsychotic medication in CHR-P and ESZ groups using Pearson correlations. In CHR-P, clinical correlates included SOPS summary scales (positive, negative, disorganized, general). In ESZ, clinical correlates included antipsychotic dosage in chlorpromazine equivalents (CPZ) (Woods, 2003) and PANSS summary scales (positive, negative, general). Significance of these correlations within each morphometric measure type were adjusted for multiple tests with false discovery rate (FDR) correction, p_FDR<.05.

3. Results

3.1. Participant demographics and clinical characteristics

See Table 1 for group demographics and clinical characteristics. Groups did not statistically differ on sex (${\chi }^2$(2, N=304) = 4.64, p=.098). However, as expected, groups differed in age (F(2,303)=9.05, p<.001), with the CHR-P group being younger than both HC (p=.011) and ESZ (p<.001) groups.

Table 1.

Demographic and Clinical Characteristics of Participant Groups

	HC Mean ± SD	CHR-P Mean ± SD	ESZ Mean ± SD
Demographics
Gender (% male)	56.56	52.81	67.74
Age (Range: 12–35)a	20.86 ± 6.36	18.68 ± 4.59	22.01 ± 4.56
Ethnicity (%)
White / Caucasian	58.19	56.18	51.61
Asian American	20.49	13.48	22.58
Black / African American	6.56	10.11	6.45
American Indian / Alaska Native	2.46	2.25	0.00
Native Hawaiian / Pacific Islander	1.64	3.37	2.15
More Than One Race	10.66	10.11	11.83
Not Reported	0.00	4.49	5.38
Medication (%)
Chlorpromazine equivalents (mg)	---	b c	285.34 ± 322.97b
Antipsychotic medication	---	14.61	80.65d
(typical, atypical, unknown)	---	(0, 100, 0)	(1.33, 94.67, 4.00)
Symptom Severity Ratings e
SOPS Positive	---	10.00 ± 4.88	---
SOPS Negative	---	11.76 ± 5.78	---
SOPS Disorganized	---	5.78 ± 3.89	---
SOPS General	---	8.07 ± 4.82	---
PANSS Positive	---	---	13.78 ± 4.70
PANSS Negative	---	---	16.55 ± 6.76
PANSS General	---	---	32.94 ± 8.49

Note. HC=healthy control; CHR-P=clinical high-risk for psychosis; ESZ=early illness schizophrenia; SD=standard deviation; SES=socioeconomic status; SOPS=Scale of Psychosis-Risk Symptoms; PANSS=Positive and Negative Syndrome Scale.

^aGroups were significantly different on age, F(2,303)=9.05, p<.001, with the CHR-P group significantly younger than HC (p=.011) and ESZ (p<.001) groups.

^bChlorpromazine equivalents for 11 CHR-P participants and 12 ESZ participants was missing.

^cMorphometric results showed the same pattern when medicated CHR-P patients were dropped.

^dAntipsychotic medication status for 7 participants missing.

^eSOPS Negative for 4 participants missing; SOPS Disorganized and General for 3 participants missing; PANSS scores only for UCSF participants.

3.2. Cortical thickness profiles

Cortical thickness ROI z-score profiles among the three groups (Figure 1 and Supplemental Table 6) showed significant non-parallelism (F[66,538]=1.86, Wilks’ $\lambda$=.66, p<.001, ${\eta }_p^2$=.19) and level differences (F[2,301]=21.83, p<.001, ${\eta }_p^2$=.13). In 2-group follow-up comparisons, the ESZ profile significantly deviated from the flat HC z-score profile (F[33,181]=2.31, Wilks’ $\lambda$=.42, p<.001, ${\eta }_p^2$=.30) and from the CHR-P profile (F[33,148]=2.03, Wilks’ $\lambda$=.69, p=.002, ${\eta }_p^2$=.31). Additionally, the ESZ group showed an overall reduction in profile level relative to HC (F[1,213]=40.68, p<.001, ${\eta }_p^2$=.16) and CHR-P (F[1,180]=25.88, p<.001, ${\eta }_p^2$=.13) groups, with the ESZ group having widespread cortical thinning across all regions except for entorhinal and pericalcarine ROIs (Supplemental Table 6). There were no significant differences between the HC and CHR-P groups (ps>.280, ${\eta }_p^2$=.00 to .18).

Figure 1.

Cortical thickness profiles of healthy control (HC), clinical high-risk (CHR), and early illness schizophrenia (ESZ) groups. All groups were nonparallel and had significant level differences. In 2-group follow-up comparisons, HC vs. ESZ and CHR-P vs. ESZ were nonparallel and had significant level differences.

3.3. Cortical surface area profiles

There was no significant non-parallelism (F[66,538]=1.24, Wilks’ $\lambda$=.75, p=.110, ${\eta }_p^2$=.13) or level difference (F[2,301]=0.40, p=.671, ${\eta }_p^2$=.00) of cortical surface area ROI z-score profiles between the HC, CHR-P, and ESZ groups (Figure 2 and Supplemental Table 7).

Figure 2.

Cortical surface area profiles of healthy control (HC), clinical high-risk (CHR), and early illness schizophrenia (ESZ) groups. Groups were parallel and had no significant level difference.

3.4. Basal ganglia volume profiles (caudate, putamen, pallidum)

Basal ganglia volume ROI z-score profiles among the three groups (Figure 3a and Supplemental Table 8) showed significant non-parallelism (F[4,600]=4.21, Wilks’ $\lambda$=.95, p=.002, ${\eta }_p^2$=.03) and level differences (F[2,301]=2.64, p<.001, ${\eta }_p^2$=.07). In 2-group follow-up comparisons, the ESZ profile significantly deviated from the flat HC z-score profile (F[2,212]=5.77, Wilks’ $\lambda$=.95, p=.004, ${\eta }_p^2$=.06) and the CHR-P profile (F[2,179]=4.32, Wilks’ $\lambda$=.95, p=.001,5 ${\eta }_p^2$=.05). Additionally, the ESZ group showed overall greater volume relative to HC (F[1,213]=11.82, p<.001, ${\eta }_p^2$=.01) and CHR-P (F[1,180]=18.861, p<.001, ${\eta }_p^2$=.10) groups, with level differences driven by enlarged caudate and pallidum volumes in ESZ (Supplemental Table 8). There were no significant differences between HC and CHR-P groups (ps>.141, ${\eta }_p^2$=.01 to .02).

Figure 3.

Subcortical/ventricular/cerebellar volume profiles of healthy control (HC), clinical high-risk (CHR), and early illness schizophrenia (ESZ) groups. a) All groups were nonparallel and had significant level differences. In 2-group follow-up comparisons, HC vs. ESZ and CHR-P vs. ESZ were nonparallel and had significant level differences. b) All groups were parallel and had no significant level differences. c) All groups were parallel and had significant level differences for ventricular volume. ESZ group had larger ventricular volumes relative to HC and CHR-P groups. d) All groups were nonparallel and had no significant level differences. In 2-group follow-up comparisons, HC vs. ESZ and HC vs. CHR were nonparallel.

3.5. Subcortical volume profiles (accumbens, hippocampus, amygdala, thalamus)

There were no significant differences in parallelism (F[6,598]=0.76, Wilks’ $\lambda$=.99, p=.600, ${\eta }_p^2$=.01) or level (F[2,301]=0.97, p=.379, ${\eta }_p^2$=.01) of subcortical volume ROI z-score profiles among the HC, CHR-P, and ESZ groups (Figure 3b and Supplemental Table 8).

3.6. Ventricular volume profiles

Ventricular volume ROI z-score profiles among the three groups (Figure 3c and Supplemental Table 8), were parallel (F[4,600]=1.96, Wilks’ $\lambda$=.97, p=.099, ${\eta }_p^2$=.01), but showed a significant level difference (F[2,301]=7.99, p<.001, ${\eta }_p^2$=.05). In post-hoc level comparisons between groups, ESZ had overall enlarged ventricles relative to HC (p<.001) and CHR-P (p=.005) groups. There was no significant group difference between HC and CHR-P groups (p=1.000).

3.7. Cerebellar volume profiles (white matter and cortex)

There was significant non-parallelism (F[2,301]=7.75, Wilks’ $\lambda$=.95, p<.001, ${\eta }_p^2$=.05) but no level difference (F[2,301]=0.64, p=.531, ${\eta }_p^2$=.00) of cerebellar volume ROI z-score profiles between the HC, CHR-P, and ESZ groups (Figure 3d and Supplemental Table 8). In 2-group follow-up comparisons, the ESZ profile significantly deviated from the flat HC z-score profile (F[1,213]=12.77, Wilks’ $\lambda$=.94, p<.001, ${\eta }_p^2$=.06) but not the CHR-P profile (F[1,180]=1.23, Wilks’ $\lambda$=.99, p=.258, ${\eta }_p^2$=.01). The ESZ did not show an overall difference in profile level relative to HC (F[1,213]=0.22, p-.642, ${\eta }_p^2$=.00) or CHR-P (F[1,180]=1.26, p=.264, ${\eta }_p^2$=.01) groups. Additionally, the CHR-P profile significantly deviated from the flat HC z-score profile (F[1,209]=7.60, Wilks’ $\lambda$=.97, p=.006, ${\eta }_p^2$=.04) but showed no level difference (F[1,209]=0.52, p=.472, ${\eta }_p^2$=.00).

3.8. Clinical high-risk for psychosis group: predicting clinical outcomes

Analyses of morphometric ROI profile differences between CHR-P non-converters and CHR-P converters (Figure 4, Supplemental Tables 9 and 10, Supplemental Figures 3 and 4) showed no significant differences in parallelism or level for cortical thickness (ps>.828, ${\eta }_p^2$=.00 to .61), cortical surface area (ps>.249, ${\eta }_p^2$=.01 to .77), basal ganglia volume (ps>.068, ${\eta }_p^2$=.00 to .11), subcortical volume (ps>.697, ${\eta }_p^2$=.00 to .32), or cerebellar volume (ps>.172 ${\eta }_p^2$=.01 to .05). For ventricular volume, there was significant non-parallelism (F[2,45]=3.49, Wilks’ $\lambda$ = .87, p=.039, ${\eta }_p^2$=.13), but no significant level difference (F[1,46]=3.17, p=.082, ${\eta }_p^2$=.06), with the CHR-P converters having larger lateral ventricles and lateral inferior ventricles (Figure 4 and Supplemental Table 10) than the CHR-P non-converters.

Figure 4.

Subcortical/ventricular/cerebellar volume profiles of clinical high-risk non-converter (CHR-NC) and clinical high-risk converter (CHR-C) groups. a and b) Groups were parallel and had no significant level difference for subcortical volume. c) Groups were nonparallel and had no significant level difference for ventricular volume. d) Groups were parallel and had no significant level difference for cerebellar volume.

3.9. Antipsychotic medication dosage associations with structural morphometry in early illness schizophrenia

To examine whether significant structural morphometric abnormalities were accounted for by antipsychotic medication, we examined associations between CPZ dosage and morphometric component z-scores in the ESZ group only (Table 2). Higher CPZ dosage was significantly associated with cortical thinning of frontal regions (r=−.36, p<.001). These associations were still significant when controlling for symptom severity (r=−.34, p=.005). There was modest evidence that higher CPZ dosage equivalents was associated with decreased cortical surface area of occipital regions (r=−.27, p=.017) and decreased subcortical volume of the amygdala/hippocampus/accumbens (r=−.22, p=.046); however, these findings did not survive FDR correction.

Table 2.

Chlorpromazine (CPZ) Equivalents and Clinical Associations with Structural Brain Morphometry in Early Illness Schizophrenia

Components	CPZ	PANSS Positive	PANSS Negative	PANSS General	Duration of Illness
Cortical Thickness
Frontal-Parietal	−.28†	.11	−.09	.10	−.08
Frontal	−.36***	.05	−.01	.16	−.15
Temporal	−.20	.07	−.05	.08	−.10
Occipital	−.14	−.12	.07	−.05	−.07
Cingulate	−.17	.15	−.21	.02	−.17
Cortical Surface Area
Frontal-Parietal	−.09	−.08	.12	−.02	−.04
Temporal	−.04	−.04	−.16	−.18	−.08
Frontal	.01	−.13	−.06	−.16	.00
Occipital	−.27†	−.03	.01	−.03	−.10
Subcortical Volume
Hippocampus/Amygdala/Accumbens	−.22†	.10	−.03	.10	−.13
Thalamus/Pallidum	.07	.00	.08	.00	.02
Caudate/Putamen	−.11	.02	.08	.10	−.18
Ventricular Volume
All Ventricles	.13	−.02	−.08	.00	.01
Cerebellar Volume
Whole Cerebellum	−.07	.03	−.02	.00	.08

Note. Bolded text denotes regions that are significant after false discovery rate correction (p_FDR<.05). PANSS=Positive and Negative Syndrome Scale. Correlations for cortical thickness adjusted for age, scanning site, and sex. Correlations for cortical surface area and subcortical/ventricular/cerebellar volume adjusted for age, scanning site, sex, and intracranial volume.

^†p<.05 but did not survive FDR correction.

^***p<.001.

3.10. Clinical associations with structural morphometry in clinical groups

In the ESZ group, PANSS scores and duration of illness were not significantly associated with cortical thickness, cortical surface area, or basal ganglia/ventricular/cerebellar volume (Table 2). In the CHR-P group, higher SOPS negative symptom scores was significantly associated with decreased subcortical volume of the hippocampus/amygdala/accumbens (r=−.32, p=.003; Table 3), but this association did not survive multiple comparison correction.

Table 3.

Chlorpromazine (CPZ) Equivalents and Clinical Symptom Severity Associations with Structural Brain Morphometry in Clinical High-Risk

	SOPS
Components	Positive	Negative	Disorganized	General
Cortical Thickness
Frontal-Parietal	.06	−.11	−.01	−.06
Frontal	.13	.02	.03	.06
Temporal	.03	−.01	−.05	.00
Occipital	−.01	−.10	.12	−.07
Cingulate	.17	.05	.08	−.04
Cortical Surface Area
Frontal-Parietal	−.11	.02	−.12	.05
Temporal	−.11	−.09	−.12	.02
Frontal	−.10	.10	.07	.10
Occipital	.01	.08	−.03	.08
Subcortical Volume
Hippocampus/Amygdala/Accumbens	−.10	−.32†	−.20	−.12
Thalamus/Pallidum	.08	−.04	.16	.07
Caudate/Putamen	−.05	.02	.12	.17
Ventricular Volume
All Ventricles	−.07	.08	−.09	.00
Cerebellar Volume
Whole Cerebellum	.02	.09	.03	.12

Note. SOPS=Scale of Psychosis-Risk Symptoms. Correlations for cortical thickness adjusted for age, scanning site, and sex. Correlations for cortical surface area and subcortical/ventricular/cerebellar volume adjusted for age, scanning site, sex, and intracranial volume.

^†p<.05 but did not survive FDR correction.

Overall, morphometric results showed the same pattern when medicated CHR-P individuals were dropped from analyses.

4. Discussion

In the current study, we examined the extent to which cortical thickness, cortical surface area, subcortical volumeventricular volume, and cerebellar volume are abnormal in CHR-P and ESZ patients relative to HC individuals after accounting for normal brain maturation. We found that ESZ individuals had widespread cortical thinning as well as increased basal ganglia and lateral ventricular volume compared to HC and CHR-P individuals, but there were no significant group differences in cortical surface area or cerebellar volume. Higher antipsychotic dosage was significantly associated with cortical frontal thinning in ESZ. In contrast to ESZ, only regional morphometric brain measure profiles significantly differed between HC and CHR-P groups for cerebellar volumes. When CHR-P converters were compared to non-converters followed for at least 18-months, CHR-P converters showed significantly larger lateral ventricles. Thus, results support the theory that schizophrenia is a progressive brain disorder, with little evidence of brain morphometric abnormalities predating psychosis onset, except for a finding larger ventricles in CHR-P individuals who subsequently converted to psychosis.

The profile of structural brain morphometric abnormalities in ESZ was consistent with literature finding widespread cortical thinning and subcortical/ventricular abnormalities in ESZ relative to HC (e.g., Akudjedu et al., 2020; Cahn et al., 2002; Crespo-Facorro et al., 2011; Fan et al., 2019; Gallardo-Ruiz et al., 2019; Lin et al., 2019; Merritt et al., 2021; Rosa et al., 2010; Zhao et al., 2022). Although some studies have found smaller cerebellar volumes in ESZ (Bottmer et al., 2005; Chan et al., 2011; Kim et al., 2018)), we did not replicate tis finding. Discrepancies in this phenomenon across studies (Bottmer et al., 2005; Chan et al., 2011; Kim et al., 2018; Morimoto et al., 2021; Steen et al., 2006) could be a result of heterogeneity in parcellating the cerebellum. Due to the relationship between structural morphometry and antipsychotic medications (Fusar-Poli et al., 2013; Ho et al., 2011; van Erp et al., 2018; Vita et al., 2015), we examined associations with antipsychotic dosage in the ESZ group. Higher CPZ dosage was associated with cortical thinning in the frontal cortex, with associations remaining when controlling for symptom severity. These findings are consistent with previous research (Cannon et al., 2016; Haijma et al., 2013). In contrast to this literature, we did not find higher CPZ dosage to be associated with basal ganglia enlargement (Ho et al., 2011; Taylor et al., 2007). It is possible that we did not find the expected association because we were only examining the effect of current antipsychotic mediation and not lifetime exposure. Although antipsychotic medication effects on structural morphometry cannot be ruled out, critically, ESZ differences were widespread and were not isolated to the frontal cortex. Previous research has also found both cortical structural deficits in ESZ individuals taking antipsychotic medications (e.g., Fan et al., 2019) and in antipsychotic-naïve ESZ individuals (Crespo-Facorro et al., 2011; Hazlett et al., 2008; Wei et al., 2022) but see (Lesh et al., 2015). Taken together, this suggests that structural morphometric differences are not merely a byproduct of antipsychotic medication, but are involved in the pathophysiology of the early course of schizophrenia.

With regard to the CHR-P group, there were no significant differences in structural morphometric profiles relative to the HC group. Although a meta-analysis found cortical thinning of the bilateral medial prefrontal cortex of CHR-P individuals relative to HC (Zhao et al., 2022), there is much variability in previous studies, with many studies finding no significant difference between groups (Bakker et al., 2016; Bois, Ronan, et al., 2015; Buechler et al., 2020; Cannon et al., 2015; Klauser et al., 2015; Ziermans et al., 2012). The largest multi-site CHR-P study to date found widespread cortical thinning in CHR-P compared to HC (Jalbrzikowski et al., 2021). However, between-group effect sizes were small (d=−0.09 to −0.17), which might explain why we did not detect abnormalities in our smaller sample CHR-P group. Interestingly, in contrast to our hypothesis of a level difference between the two clinical groups, morphometric CHR-P profiles were not just indicative of reduced levels of abnormalities present in ESZ, but profiles were also nonparallel. Given that CHR-P individuals are a heterogeneous group, with the majority not converting to a psychotic disorder (Cannon et al., 2008; Fusar-Poli et al., 2012; Simon et al., 2014), it is possible that this heterogeneity is resulting in both level differences and nonparallelism between ESZ and CHR-P. Subtle morphometric changes in the high-risk period may be more apparent and show a more similar pattern to the early illness phase in CHR-P individuals who subsequently develop a psychotic disorder.

Consistent with this idea, studies have found baseline cortical thinning (Buechler et al., 2020; del Re et al., 2021; Jalbrzikowski et al., 2021) and steeper rates of cortical thinning (Cannon et al., 2015) as well as ventricular expansion (Cannon et al., 2015; Chung et al., 2017; Sasabayashi et al., 2020) in CHR-P individuals who convert to a psychotic disorder versus those who do not. In the current study, our CHR-P converter group showed enlargement of lateral and inferolateral ventricles. Null cortical thickness findings should be interpreted with caution because of the low statistical power due to small samples of CHR-P converters and CHR-P non-converters followed for at least 18-months.

Lastly, structural morphometry was not significantly associated with clinical symptoms in the two clinical groups or duration of illness in ESZ. Research on structural morphometric associations with clinical symptoms in the high-risk period and early illness phase is limited. Of the few CHR-P studies that have examined clinical associations, most have not found them at all (Bakker et al., 2016; Buechler et al., 2020; del Re et al., 2021; Jung et al., 2011) or reported only weak associations (Kwak et al., 2019). For the early illness phase, Crespo-Faccoro et al. (2011) found no significant clinical symptom associations with cortical regions, but Fan et al. (2019) found PANSS positive symptoms to be inversely related to volumes of the amygdala and nucleus accumbens. Further, previous research in chronic schizophrenia patients has found evidence that greater illness duration is associated with cortical thinning of the insula (van Erp et al., 2018) and increased putamen and pallidum volume (van Erp et al., 2016), but not with cortical surface area (van Erp et al., 2018). Correlations in the current study could be attenuated by the restricted range of illness duration inherent in an ESZ sample.

Some current study limitations should be noted. Evidence of cortical thinning at baseline is mixed in CHR-P, with some longitudinal studies finding steeper rates of cortical thinning in CHR-P who later convert to psychosis compared to HC (Cannon et al., 2015). As such, longitudinal designs might be more sensitive to the subtle progressive structural changes developing in this group. Additionally, the current study was limited by insufficient follow-up CHR-P data and a relatively small CHR-P conversion group. Due to previous evidence of cortical thinning in CHR-P converters relative to CHR-P non-converters (Buechler et al., 2020; del Re et al., 2021; Jalbrzikowski et al., 2021), future studies should include larger samples in order to conduct sufficiently powered conversion analyses (Jalbrzikowski et al., 2021). Future studies should also examine how structural morphometry is related to other functional brain measures assessed with methods like functional MRI and electroencephalography measures (e.g., Hua et al., 2022).

In conclusion, the current study found widespread cortical thinning and subcortical/ventricular volumetric enlargement in the early illness phase of schizophrenia, but not in individuals at clinical high-risk for psychosis. Ventricle enlargement in the CHR-P group was only apparent in CHR-P individuals who converted to psychosis. By concurrently comparing structural morphometric profiles of HC, CHR-P, and ESZ groups, the current study extends previous morphometric research and found novel evidence that the CHR-P structural morphometric profile is not just a reduced level of the ESZ profile, but is also nonparallel to the ESZ profile. Although more longitudinal studies are needed to examine morphometric progression, results are consistent with the view that schizophrenia is a progressive brain disorder, with abnormalities being significantly worse during the early years of schizophrenia relative to those exhibiting symptoms of a psychosis-risk syndrome.

Highlights.

• Compared clinical high-risk (CHR-P) and early illness schizophrenia (ESZ) profiles

• ESZ morphometric profiles differed from healthy controls (HC) and CHR-P

• Only cerebellar morphometric profile differences emerged between CHR-P and HC

• CHR-P converters vs. non-converters showed regional ventricular volume expansion

• Antipsychotic dosage in ESZ was correlated with frontal cortical thinning