Cortical Thickness Abnormalities at Different Stages of the Illness Course in Schizophrenia

1Schizophrenia (SCZ) is a debilitating psychotic disorder typically beginning in late adolescence or early adulthood, often followed by recurrent episodes of psychosis and, in some cases, declining psychosocial function. The onset of illness is usually preceded by a clinical high-risk (CHR) phase characterized by attenuated psychosis symptoms, brief limited psychotic symptoms, and functional deterioration. Potential brain changes between CHR and onset of the first episode of psychosis (FEP) that might reflect neurotoxicity of acute psychosis, and between FEP and long-term illness that might reflect progressive brain abnormalities, are important issues being addressed in several SCZ research programs.

2Previous neuroimaging studies have documented widespread structural brain alterations at different stages of SCZ., However, because most studies focused on only 1 specific illness stage, questions of how these alterations may differ over the illness course have yet to be fully resolved. Few longitudinal studies have been completed with currently available image analysis pipelines, and few studies have compared patients with CHR and FEP. Some cross-sectional studies of patients after illness onset have identified an accelerated rate of age-related decline in brain features compared with healthy individuals, which provides preliminary evidence for progressive anatomic changes over the illness course.,,, A systematic review and meta-analysis comparing individuals at different illness stages might characterize brain abnormalities over the illness course with robustness enhanced by combining data across the published literature.

3Clinically, there are 2 broad issues of particular interest. The first is whether there are greater brain abnormalities in individuals with FEP than CHR. This issue is related to questions about whether the onset of acute psychosis itself induces brain alterations. The second issue is whether, following onset of the illness, there are increasing brain alterations that might be targets for therapeutic intervention. Addressing these issues is also important from the perspective of brain imaging biomarker development.

4Recently, regional cortical thickness (CTh) measurement via surface-based morphometry has been used to evaluate brain anatomy in patients with SCZ. Cortical thickness analysis can better reflect underlying pathophysiologic mechanisms than gray matter volume analysis. Cortical thickness alterations have been identified in SCZ, but the regional distribution of detected abnormalities has not been consistent, even within studies of specific phases of illness. Individuals with CHR have demonstrated varying CTh alterations,,, with those who subsequently developed psychosis having greater abnormalities.,, Cortical thinning has been reported in patients with FEP in prefrontal, lateral temporal,, and parietal cortex, albeit with negative findings., Discrepancies might be related to small sample sizes, variable sample characteristics, medication status, and analytical methods.

5Few systematic review and meta-analysis studies have examined CTh, with most considering volume measures (eTable 1 in the Supplement),, and none compared thickness across the 3 illness stages. Findings from earlier studies include a smaller parahippocampal volume in long-term SCZ than FEP. Both individuals with CHR and those with FEP have shown similar cortical thickening in bilateral occipital regions. Individuals with FEP exhibited thinner right posterior cingulate cortex than those with CHR. More widespread and pronounced CTh reductions were found in individuals with long-term SCZ than CHR and FEP. A limited number of longitudinal studies have found progressive CTh reductions in long-term SCZ, predominantly in frontotemporal regions. These findings generally parallel observations from neurophysiologic and neuropsychological studies.,

6Given the clinical importance of understanding the course of illness in SCZ and the variability in findings across single studies, we conducted coordinate-based meta-analyses of published studies on CTh in individuals with CHR, FEP, and long-term SCZ, using seed-based d mapping software. We performed case-control meta-analyses in each illness phase, followed by quantitative comparisons of findings between illness stages. Our primary interest was testing for increases in abnormalities in FEP compared with CHR samples and in long-term SCZ compared with FEP samples. We tested for associations between clinical and demographic variables with anatomic alterations, using meta-regression analyses.

7Systematic review and meta-analysis was carried out using the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline. The protocol was registered on PROSPERO (CRD42021253875). Details of the search strategy and selection criteria are provided in the eMethods in the Supplement. Two of us (Y.Z. and Q.Z.) independently conducted literature searches in PubMed, Embase, Web of Science, and Science Direct and screened the results for CTh studies published before June 15, 2021. Inconsistencies were discussed and a consensus was reached.

8Meta-analyses of CTh studies were conducted using seed-based d mapping software (SDM, version 5.15, SDM Project). First, coordinates of cluster peaks and effect sizes of significant between-group differences were used to create an effect-size signed map for each study. Negative findings were included and estimated conservatively to have a null effect size. Next, a random-effects analysis was performed to obtain the mean map, combining data of each included study with both positive and negative differences in the same map.

9Separate analyses were performed to identify CTh alterations at each illness stage vs healthy control individuals. We then performed a pooled meta-analysis comparing patients at different illness stages using randomization tests to identify significant differences and a regression analysis across all samples to test for correlations with demographic and clinical features. Quadratic models of age effects were used when significantly better than linear models. Details of quality assessment and data recording, seed-based d mapping method of meta-analysis, jackknife, heterogeneity, and publication bias analysis, and meta-regression analysis are presented in the eMethods in the Supplement. All tests were 2-sided and unpaired, and a statistical threshold of P <.005 was used with a cluster extent of 10 voxels for the meta-analysis.

10Figure 1 shows the PRISMA flowchart of the literature search and eligibility assessment. Ten studies,,,,,,,,, comprising 859 individuals with CHR (mean [SD] age, 21.02 [2.66] years; male, 573 [66.7%]; female, 286 [33.3%]) and 531 healthy control individuals (mean [SD] age, 21.67 [3.03] years; male, 300 [56.5%]; female, 231 [43.5%]), 12 studies,,,,,,,,,,, including 671 individuals with FEP (mean [SD] age, 22.87 [3.99] years; male, 439 [65.4%]; female, 232 [34.6%]) and 566 healthy control individuals (mean [SD] age, 23.39 [4.5] years; male, 349 [61.7%]; female, 217 [38.3%]), and 10 studies,,,,,,,,, comprising 579 individuals with long-term SCZ (mean [SD] age, 41.58 [6.95] years; male, 396 [68.4%]; female, 183 [31.6%]) and 475 healthy control individuals (mean [SD] age, 40.07 [7.0] years; male, 303 [63.8%]; female, 172 [36.2%]) were included in the meta-analysis. Four of 12 FEP studies included some patients with a current diagnosis of affective psychosis (proportion, 9.7%). For longitudinal studies, only baseline data were included to avoid bias toward the outcomes of the interventions or illness progression. Results of quality assessments (eResults and eTable 2 in the Supplement) and detailed characteristics of included studies (eResults and eTable 3 in the Supplement) are provided.

11Compared with healthy control individuals, individuals with CHR showed cortical thinning of 1 cluster in bilateral medial prefrontal cortex (z = −1.01; P <.001) extending to bilateral superior frontal cortex, bilateral anterior cingulate cortex, and right middle cingulate cortex (Figure 2, Table; eFigure 1 in the Supplement). No significant association of age was found in this cluster (eTable 4 in the Supplement).

12Individuals with FEP showed CTh reductions in 3 clusters: right lateral superior temporal cortex (z = −1.34; P<.001), right anterior cingulate cortex (z = −1.44; P <.001), and right insula (z = −1.14; P =.002) (Figure 2, Table; eFigure 2 in the Supplement). No significant association of age or illness duration was found in any cluster (eTable 4 in the Supplement).

13Individuals with long-term SCZ showed thinner cortex than healthy control individuals in 4 clusters: right insula extending to frontal operculum (z = −2.58; P <.001), right pars orbitalis of inferior frontal cortex (z = −2.32; P <.001), left lateral temporal cortex (z = −1.91; P =.002), and right temporal pole (z = −1.82; P =.002) (Figure 2, Table; eFigure 3 in the Supplement). In individuals with long-term SCZ, illness duration was associated with greater CTh reductions only in left anterior temporal cortex (R2 = 0.60; P =.003) (eFigure 4, eTable 4 in the Supplement).

14We found no significant CTh differences between the CHR and FEP groups. Individuals with long-term SCZ showed greater cortical thickness reductions in right insula, right pars orbitalis of inferior frontal cortex, left lateral temporal cortex (including superior temporal cortex and middle temporal cortex [MTC]), and right temporal pole than those with FEP (Figure 3). Similar CTh reductions were seen in individuals with long-term SCZ compared with CHR (eFigure 5, eTable 5 in the Supplement).

15The combined group (all 3 illness stages) showed cortical thinning in 5 clusters, including right insula extending to frontal operculum, left anterior cingulate cortex extending to bilateral middle cingulate cortex, right pars orbitalis of inferior frontal cortex, and bilateral lateral MTC (eFigure 6 in the Supplement; Table).

16Combining data sets across illness stages, reduced CTh in right pars orbitalis of inferior frontal cortex compared with healthy control individuals was correlated with older age (quadratic model: R2 = 0.28; P =.004) (Figure 4A). Cortical thickness reductions in individuals with left lateral MTC compared with healthy control individuals were correlated with greater age (linear model: R2 = 0.47; P <.001; quadratic model: R2 = 0.64; P <.001) (Figure 4B) and greater illness duration (linear model: R2 = 0.50; P <.001) (eFigure 7 in the Supplement). In the homologous right hemisphere, reduced CTh in right lateral MTC also was correlated with older age (linear model: R2 = 0.34; P <.001; quadratic model: R2 = 0.44; P <.001) (Figure 4C; eTable 6 in the Supplement). The quadratic model had a significantly lower root-mean-square error value than linear model in left lateral MTC (quadratic model: median [range], 0.12 [0.09-0.12]; linear model: median [range], 0.14 [0.12-0.14]; z = 5.37; P <.001) and right lateral MTC (quadratic model: median [range], 0.13 [0.09-0.13]; linear model: median [range], 0.14 [0.11-0.14]; z = 5.37; P <.001) (eTable 7 in the Supplement), indicating better performance of the quadratic model than the linear model in both regions. The findings for onset age or positive, negative, or general symptom ratings on CTh were not statistically significant in the combined group (eTable 6 in the Supplement).

17Whole-brain jackknife sensitivity analysis confirmed replicability and reliability of the findings (eResults 3, eTables 8-11 in the Supplement). Cortical thickness reductions in right insula showed high heterogeneity (I2 = 82.50%) in FEP groups and moderate heterogeneity (I2 = 45.00%) in the combined group, raising questions about the validity of findings in this region, so they were not considered further. The remaining regions with altered CTh showed low heterogeneity (I2 <15.00%) between studies (eResults in the Supplement). The Egger test of funnel plot asymmetry was not statistically significant when testing for publication bias (eResults, eFigures 8-11 in the Supplement).

18Coordinate-based meta-analyses of CTh abnormalities across illness stages in SCZ identified robust CTh reductions in individuals with CHR, FEP, and long-term SCZ. There were no significant CTh differences between those with CHR and FEP. In contrast, alterations in CTh of individuals with long-term SCZ were more pronounced than seen in FEP studies in right inferior frontal cortex, left lateral temporal cortex (including superior temporal cortex and MTC), and right temporal pole. When combining studies regardless of illness stage, we observed accelerated age-related reductions of CTh in bilateral lateral MTC and right pars orbitalis compared with healthy control individuals. These findings did not demonstrate a robust reduction of CTh at illness onset but suggest a progressive age-related reduction in CTh later in the illness course.

19Individuals with CHR showed cortical thinning in bilateral medial prefrontal cortex, which plays an important role in cognitive functioning, emotion regulation, and motivation. Previous functional evidence suggested that hypoactivation of medial prefrontal cortex is related to anhedonia and poor episodic and working memory in SCZ, which are cognitive impairments already evident in individuals with CHR., Our findings were consistent with an earlier systematic review and meta-analysis that identified reduced medial prefrontal cortex volume in individuals with CHR. Moreover, greater gray matter loss in medial prefrontal cortex has been found in individuals with CHR who subsequently developed psychosis. Thus, our findings and previous volumetric studies converge in suggesting that anatomic alterations in medial prefrontal cortex precede the illness onset and thus may provide a marker of vulnerability to psychosis in individuals with CHR.,

20Individuals with FEP demonstrated CTh reductions in right lateral superior temporal cortex and right anterior cingulate cortex relative to healthy control individuals. The superior temporal cortex contains primary and secondary auditory cortices and supports language comprehension. Decreased CTh, reduced gray matter volume, hypoactivation of superior temporal cortex, and reduced functional connectivity between superior temporal cortex and parietal areas have been associated with auditory verbal hallucinations, disordered thinking, and impaired speech recognition in SCZ. The anterior cingulate cortex is a major hub of the salience network engaged in cognitive, homeostatic, motivational, and affective functions. Gray matter loss and cortical thinning in this region have been linked to auditory hallucinations in FEP.

21In no brain region did the severity of CTh alterations differ significantly between individuals with CHR and FEP. This lack of difference was seen when all individuals with CHR were considered—not just those who went on to manifest psychosis. This finding suggests that the CTh alterations seen in both groups for the most part reflect either a predisposing or neurodevelopmental feature already present by the time of prodrome manifestations. Furthermore, the lack of greater deficits in the FEP than CHR studies suggests that the onset of acute psychosis itself does not have robust neurotoxic outcomes associated with CTh.

22Individuals with long-term SCZ exhibited cortical thinning in right inferior frontal cortex and bilateral temporal pole relative to healthy control individuals, in accordance with previous reports showing gray matter loss in these regions in individuals with long-term SCZ. The inferior frontal cortex is critical for affect modulation and empathy, which is impaired in SCZ. The temporal pole is part of the extended paralimbic system involved in emotion processing and interoceptive awareness. Gray matter volume reductions in paralimbic regions (including orbitofrontal cortex and temporal pole) have been related previously to cognitive and affective dysfunctions and clinical symptom severity in long-term SCZ.

23Individuals with long-term SCZ showed more pronounced CTh reductions in left lateral temporal cortex, right temporal pole, and right inferior frontal cortex than were seen in FEP studies considered relative to their own age-matched controls. These group differences suggest a progressive change in brain anatomy in SCZ relative to healthy control individuals over the years after illness onset. Such findings are of potential clinical importance, as greater gray matter volume and CTh reductions in frontotemporal regions have been associated with deteriorated cognitive deficits and poorer clinical outcomes that are a common and primary source of disability in this illness.

24Consistent with the possibility that differences between individuals with FEP and those with long-term SCZ may reflect aspects of disease progression, when pooling all included studies regardless of illness stage, we found accelerated age-related CTh reductions in bilateral lateral MTC and right pars orbitalis of inferior frontal cortex. These observations support the general idea of neuroprogression, but one localized to frontotemporal regions. Although the source of these progressive changes needs to be determined, because they may be directly related to illness pathophysiologic changes or be secondary to antipsychotic treatment effects or other environmental factors, the changes may themselves reflect a target for treatment development to reduce the longer-term morbidity associated with the disorder. The accelerated age-related CTh loss in frontotemporal cortex was often better modeled with a quadratic than linear model, suggesting accelerating progressive alterations in this region. Cortical thickness reductions of bilateral lateral MTC and right inferior frontal cortex appeared to accelerate in midlife, starting in the 40s, in accordance with previous volumetric studies showing steeper age-related gray matter volume loss in the frontotemporal regions.

25This study has limitations. First, although our meta-analytic approach provides the opportunity to combine cross-sectional studies across the literature to evaluate a potential regional pattern of progressive atrophic changes, longitudinal studies for confirmation may be warranted. Second, at first presentation of acute psychosis, differential diagnosis of affective and nonaffective disorders is challenging. Four of 12 FEP studies included patients with a current diagnosis of affective psychosis (with a proportion no more than 10%) (eTable 12 in the Supplement). This overlap in illness risk may also be relevant for CHR samples but could affect comparisons of FEP and long-term illness studies. Third, we were unable to explore medication therapy because precise information about dosage was not always available. Antipsychotic medications may contribute to the group differences and age outcomes in the long-term SCZ group. Fourth, the neuropathologic mechanism of cortical thinning has yet to be elucidated. Postmortem studies are needed to clarify the neural alterations leading to the findings observed in MRI studies. Such information could guide development of novel interventions to reduce these midlife brain changes in patients with SCZ. Fifth, our analyses focused on the thickness of cortical gray matter. Other anatomic features of neocortex merit attention in future studies, as does subcortical gray matter. Sixth, there is considerable heterogeneity in individuals meeting criteria for CHR that is yet to be resolved. Some will develop psychosis but most will not, and some will develop illnesses other than SCZ. Identifying discrete heterogeneity in individuals with CHR in association with brain features and clinical outcomes may ultimately lead to a better understanding of the similarities and differences of subgroups of patients with CHR compared with those who have FEP. We seem to be a few years from that point, as developing large samples and monitoring them longitudinally is an extensive effort, but examining discrete heterogeneity will greatly contribute to understanding of the CHR group. Seventh, despite documentation of limited interstudy heterogeneity in all regions but the insula, a small number of studies might influence regional study findings in our systematic review and meta-analysis. The variability across studies, although considered and evaluated statistically, raises caution in interpreting our systematic review and meta-analysis findings. This possibility will need to be evaluated as more published studies become available.

26The findings of this coordinate-based systematic review and meta-analysis identified CTh reductions in individuals with CHR that did not differ significantly from those observed in FEP studies. In addition, we observed CTh reductions in individuals with long-term SCZ compared with FEP that increased with age in frontotemporal regions and may reflect a neuroprogressive process. The lack of significant differences between individuals with CHR and FEP does not provide evidence indicating that first manifestation of acute psychosis is associated with robust CTh reductions. Together, these findings support understanding of the neural substrates of SCZ manifested at specific stages of illness and expand characterization of increasing illness-related neuroanatomic alterations over the illness course.