Patterns of Brain Structural Changes in First-Contact, Antipsychotic Drug-Naïve Patients with Schizophrenia

BACKGROUND AND PURPOSE: Previous studies have suggested that structural changes do occur in the brain of patients with schizophrenia compared with healthy control participants. However, findings from such studies are inconclusive, probably because of the different methodologic approaches, the clinical heterogeneity of patient samples, and also the fact that patients enrolled were treated with antipsychotic drugs. The aim of this study was to investigate brain GM volumes and intrinsic structural WM changes in first-contact, antipsychotic drug-naïve patients with schizophrenia. MATERIALS AND METHODS: A total of 43 first-contact, drug-naïve, patients with schizophrenia and 17 age-matched control participants were studied. All participants underwent T1-weighted MR imaging and DTI scans. Voxel-based morphometry and tract-based spatial statistics were used to compare GM volumes and WM DTI metrics between groups. MR imaging measures were correlated with the duration of the untreated psychosis and the clinical positive and negative symptoms. RESULTS: Compared with control participants, patients with schizophrenia showed smaller volumes of the temporal, parietal, and occipital GM, and a pattern of decreased mean diffusivity and increased fractional anisotropy in the brain stem and cerebellum bilaterally, interhemispheric and cortico-cortical connections bilaterally, and right anterior and posterior limb of the internal capsule. In patients, decreased mean diffusivity and increased fractional anisotropy in several brain regions were related to a longer duration of the untreated psychosis and the severity of positive symptoms. CONCLUSIONS: First-contact, drug-naïve, patients with schizophrenia present with volumetric and DTI changes, which correlated with their clinical features. This study increases our knowledge on the neural networks involved in the pathophysiologic mechanisms of schizophrenia.

I n recent years, interest in the physiopathologic mechanisms of schizophrenia has increased dramatically. As normal brain functions are served by networks of cortical and subcortical areas, disturbed communication ("dysconnectivity") within and between brain regions may be the core pathologic feature of schizophrenia. 1 In this context, the contribution of neuroimaging techniques has proved to be relevant. Structural MR imaging studies have shown that reductions of GM volume occur in the brain of patients with schizophrenia compared with healthy control participants, though increased GM volumes of some subcortical regions have also been reported. 2 DTI studies of schizophrenic patients have disclosed an abnormal organization and integrity of several WM tracts of the brain. However, results are conflicting regarding which tracts are affected, with some DTI studies showing a marked involvement of the anterior pattern of brain WM regions in patients with schizophrenia, and others describing a posterior pattern of WM changes or even no difference compared with control participants. 3 The inconclusive findings of previous studies are probably the result of the clinical heterogeneity of the samples and/or the dif-ferent MR imaging approaches used. In addition, most previous studies have enrolled medicated patients, and the effects of antipsychotic drugs and treatment duration on MR imaging findings were not taken into account. It is clear that studies based on untreated patients enrolled at an early stage of the disease represent the easiest and most rewarding strategy to overcome such an issue. To date, however, only a few studies [4][5][6][7][8][9] have adopted this strategy showing WM changes compared with healthy participants in a variable pattern of brain regions including the splenium of the corpus callosum, 4,6 superior longitudinal fasciculus, 7,8 anterior 9 and posterior 4 limbs of the internal capsule, and fronto-occipital fasciculus. 4 The relevance of DTI abnormalities in the early phase of schizophrenia is further supported by the positive association between positive symptoms and fractional anisotropy (FA) values of the superior and inferior longitudinal fasciculi and inferior fronto-occipital fasciculus in never-medicated patients with schizophrenia. 5,10 In addition, most studies 4-8 have investigated FA only, and none of them have assessed GM volumes in the same sample.
The aim of this study was to investigate whether volumetric and DTI changes, in FA, mean diffusivity (MD), and axial and radial diffusivity, are present in a relatively large sample of firstin-lifetime psychiatric contact, antipsychotic drug-naïve patients with a diagnosis of schizophrenia, compared with control participants, and to explore whether brain structural changes in these patients are associated with the duration of the untreated psychosis (DUP) and the severity of clinical positive and negative symptoms. We hypothesized that antipsychotic drug-naïve patients with schizophrenia show WM changes in cortical and subcortical connection systems, in both increased and decreased diffusion, and that such WM changes reflect the DUP and the severity of clinical symptoms. We also postulated that patients with schizophrenia in the early stage of the disease have GM volume loss in regions related to visual and episodic memory recall processing (ie, the occipital, temporal, and parietal lobes).

Participants
Patients were recruited consecutively at the Department of Psychiatry, University School of Medicine, Brescia, Italy. Control participants were selected among the personnel of the Spedali Civili of Brescia and students of the University Brescia.
To be eligible, patients and control participants had to meet the following criteria: age between 18 and 50 years; antipsychotic drug-naïve and lack of exposure to other classes of psychopharmacologic agents for more than 2 consecutive weeks preceding the enrollment; no family ties with other enrolled participants; a negative history of seizures and head trauma with loss of consciousness; no concomitant major medical conditions; Mini-Mental State Examination 11 score Ͼ 24; an IQ score on the Wechsler Adult Intelligence Scale-Revised 12 Ͼ 75; and no other causes of focal or diffuse brain damage, including lacunae, and extensive cerebrovascular disorders at routine MR imaging. Furthermore, patients were included if they had a current Diagnostic and Statistical Manual of Mental Disorders, (DSM-IV), diagnosis of schizophrenia 13 ; no lifetime comorbidities with DSM-IV, Axis I disorders; a first-in-lifetime psychiatric contact; a level of understanding judged sufficient to give informed consent; and willingness to undergo an MR imaging examination before starting an antipsychotic drug. The diagnosis of schizophrenia and the exclusion of Axis I comorbidities were based on a detailed clinical interview complemented by a revision of all medical records.
For control participants, the group-specific inclusion/exclusion criteria were no lifetime evidence of any DSM-IV, Axis I disorder; a negative family history for psychosis and mood disorders in first-degree relatives; and willingness to perform an MR imaging examination.
A total of 43 patients with schizophrenia and 17 control participants were enrolled (Table 1). After complete description of the study to the participants, written informed consent was obtained. Participants also received an explicit guarantee of anonymity. 14

Clinical Assessment
Participants were evaluated by a team of qualified and experienced psychiatrists, blinded to MR imaging data. Direct information from the patients, together with systematic interviews of at least 1 family member, was collected to date the age of onset of schizophrenia, operationally identified as when the first psychotic symptom with deterioration of function emerged. The interval elapsing between the first symptom and the first psychiatric contact defined the DUP. 15 The Positive and Negative Syndrome Scales (PANSS) 16 were administered to the patients.

MR Imaging Study
Within 2 weeks from the first clinical visit, brain MR imaging scans were obtained by use of 1.5T scanner. A detailed description of the MR imaging study is provided in the On-line Appendix. Tract-based spatial statistics (TBSS) was used to perform the multisubject DTI analysis. Voxel-based morphometry was run to assess GM volumes, by use of Statistical Parametric Mapping 8 (Wellcome Department of Imaging Neuroscience, London, UK).

Statistical Analysis
Demographic and clinical variables were compared between groups by use of the Fisher exact test, the Mann-Whitney U test, or the Kruskal-Wallis test as appropriate (P Ͻ .05; SPSS 13.0; IBM, Armonk, New York). DTI voxelwise statistics were performed by use of a permutation-based inference tool for nonparametric statistical thresholding. 17 MD, FA, axial, and radial diffusivity values within the skeleton were tested between groups by use of a permutation-based 2-sample t test adjusted for age. 17 Statistical maps were thresholded at P Ͻ .05 uncorrected for multiple comparisons.
Analyses of covariance were performed to assess GM volume differences between groups, with adjustment for age and total intracranial volume. Results were tested at P Ͻ .001 uncorrected, within at least 50 contiguous voxels. Furthermore, maps of the average percentage GM tissue loss in patients vs control participants were computed on the basis of the ratio, at each GM point, between the mean GM attenuation value at that point in the patient group and the corresponding mean GM attenuation value of the control group. The mean percentage GM tissue loss of clusters showing a significant between-group difference was also measured.
In patients, the associations of DUP and PANSS scores with DTI variables and GM volumes were tested by regression models in the FMRIB Software Library (http://www.fmrib.ox.ac.uk/fsl) and Statistical Parametric Mapping 8, respectively. Results were assessed at P Ͻ .05 for WM and P Ͻ .001 within 50 voxels for GM, uncorrected. The mean FA and MD values of the skeletonized voxels showing a significant correlation with clinical features were calculated. Then, mean FA and MD values were correlated with DUP and PANSS by use of the Pearson coefficient adjusted for the participant age (P Ͻ .05; SPSS).

Demographic and Clinical Features
Compared with control participants, patients had similar age and sex but differed in length of education (Table 1).

WM Damage: TBSS
Patients compared with control participants showed an increased MD in the fornix and thalamic radiations bilaterally, and right olfactory bulb (Fig 1). In patients vs control participants, decreased MD was found in the posterior cerebellar lobe bilaterally, cerebral peduncles, pons and medulla oblongata, middle and inferior cerebellar peduncles, genu of the corpus callosum, middle cingulum, uncinate and parahippocampal tracts bilaterally, and right anterior and posterior limbs of the internal capsule, right thalamic radiations, right inferior fronto-occipital fasciculus, and right superior and inferior longitudinal fasciculi (Fig 1). Compared with control participants, patients had a decreased FA in the cerebral peduncles, thalamic radiations, fornix, corona radiata (with a left-side predominance), corpus callosum, superior longitudinal fasciculus (with a left-side predominance) bilaterally, right superior cerebellar peduncle and posterior cerebellar lobe, and left olfactory bulb, left posterior parahippocampal tract, and left inferior fronto-occipital fasciculus (Fig 1). Compared with control participants, patients also had regions of increased FA in the cerebral peduncles, pons and medulla oblongata, middle and superior cerebellar peduncles bilaterally, and right posterior limb of the internal capsule, right external capsule, right frontoparietal part of the superior longitudinal fasciculus, and right parahippocampal tract (Fig 1).
Axial and radial diffusivity patterns resembled those of MD and FA in all contrasts. In all comparisons, patterns of decreased axial diffusivity mirrored those of reduced MD, while patterns of increased axial diffusivity reflected those of increased MD and decreased FA (On-line Figure). Patterns of decreased radial diffusivity reflected those of increased FA, whereas increased radial diffusivity reflected those of increased MD (On-line Figure).

GM Volumes: Voxel-Based Morphometry
Compared with control participants, patients showed smaller GM volumes of the bilateral postcentral gyrus; right fusiform gyrus and middle cingulate cortex; and left middle temporal and superior occipital gyrus, precuneus, and cerebellum (Table 2, Fig 2). In these regions, patients with schizophrenia had up to 20% GM loss relative to healthy control participants. No regions of increased GM volumes were found in patients compared with control participants.

Relationship between Clinical Features and GM and WM Damage
DUP. In patients, a negative relationship was found between DUP and MD values of the cerebral peduncles, pons and medulla oblongata, middle cingulum, parahippocampal tracts and inferior longitudinal fasciculi bilaterally, and right posterior limb of  the internal capsule, right thalamic radiations, right inferior fronto-occipital, and right uncinate fasciculi (Fig 3A). In patients, a positive relationship was found between DUP and FA values of the posterior limb of the internal capsule, superior longitudinal (with a left-side predominance) and inferior fronto-occipital fasciculi bilaterally, and the right uncinate fasciculi ( Fig 3A). Pearson correlation coefficients were Ϫ0.81 for MD values and 0.82 for FA values (P Ͻ .001).
PANSS Scores. In patients with schizophrenia, MD values of the right middle and inferior cerebellar peduncles, uncinate and parahippocampal tracts, inferior longitudinal fasciculus, posterior limb of the internal capsule, inferior fronto-occipital and superior longitudinal fasciculi, and the left posterior cerebellar lobe were inversely related with the PANSS positive score (Fig 3B). The Pearson correlation coefficient was Ϫ0.69 (P Ͻ .001).

DISCUSSION
To our knowledge, this study is the first to investigate both microstructural WM and volumetric GM changes in a large group of first-contact, antipsychotic drug-naïve patients with schizophrenia. We observed that the brain of patients with schizophrenia is characterized by the coexistence of WM microstructural changes and reduced GM volumes in several brain regions since the clinical onset of the disorder. The inclusion of patients at the first psychotic event further highlights the precocity of the observed brain changes. Compared with control participants, patients with schizophrenia showed WM changes, in both higher and lower diffusion, of the following networks: the limbic system, the interhemispheric connections, the cortico-cortical systems, the motor system, and the cerebello-thalamo-cortical circuit. Patients with schizophrenia also had reduced GM volume mainly of the posterior brain regions. MR imaging changes observed in our patients may be the structural correlates of the "hypoconnectivity" and "hyperconnectivity" between brain regions that have been proposed to characterize cortical and subcortical connections in schizophrenia. 18 The potential role of structural brain "disconnectivity" 18 in determination of the clinical picture of schizophrenia is further suggested by the correlation analysis. Indeed, in patients with schizophrenia, we observed a positive correlation between DUP and FA, and a negative relationship of DUP and positive symptoms with MD values of the motor and the cortico-cortical systems, suggesting the dysfunctional nature of the structural "hyperconnectivity" of these systems in untreated patients.

WM Findings
Compared with control participants, patients with schizophrenia showed decreased FA and increased MD in several regions corresponding to those observed in other studies that investigated never-medicated schizophrenic patients between 18 and 45 years old. [4][5][6][7][8] Previous authors have found lower FA values in the fornix, 7 thalamic radiations, 8   Results are overlaid on the sagittal, coronal, and axial sections of the MNI standard brain in neurologic convention (right is right) and are displayed at permutation-based P Ͻ .05, uncorrected.
prehensive review of WM findings in schizophrenia, please see Kyriakopoulos and Frangou et al 19 and Peters et al 1 ). Postmortem studies have identified WM alterations in schizophrenia, which include microstructural abnormalities of the myelin and oligodendrocytes that may represent the pathologic substrate of the observed MD increases and FA decreases in the MR imaging results of our study and in previous MR imaging reports. 20 Different from other studies of first-episode antipsychotic drug-naïve patients with schizophrenia, [4][5][6][7][8][9] we observed that patients also showed regions of increased FA and decreased MD values, likely reflecting structural "hyperconnectivity," 18 mainly in the brain stem, cerebellum, and in the interhemispheric and cortico-cortical WM connections. Some studies of patients with schizophrenia have found a similar pattern of DTI changes. 21,22 It is interesting to note that a recent study detected both increased and decreased FA in untreated patients with schizophrenia in the corticopontine-cerebellar circuit, 21 with the most inferior sections of the corticopontine tracts and the middle cerebellar peduncles showing higher FA values in patients compared with control participants (ie, exactly the same pattern that we observed in our present study; see the sagittal sections of FA:ϫ-1 and MD:ϫ5, Fig 1). Another DTI study of 12 first-episode patients with schizophrenia found an increased intervoxel coherence, which corresponds to high FA, in the thalamic radiations. 23 The pathophysiologic correlates of decreased MD and increased FA in patients with schizophrenia can be diverse. First, an increased WM fiber attenuation in patients with schizophrenia vs healthy participants is observed, as shown by postmortem studies investigating the WM architecture of patients with schizophrenia. 24 Second, there is decreased integrity of the corticopontine inhibitory GABAergic projections that may result in an increased axonal remyelination downstream, 21 which, in turn, may cause a paradoxic decrease of MD and increase of FA as observed in our study. Third, an alternative, but not mutually exclusive, explanation may be related to the dopamine hypothesis of schizophrenia, 25 according to which a hyperdopaminergic status could lead to "overdemanding" neuronal activity able to initiate an increase of useless WM connections and, clinically, induce aberrant assignments of motivational salience to objects, people, and actions.
In patients with schizophrenia, we observed that the DUP correlated negatively with MD values and positively with FA values. In addition, we also found a negative correlation between MD of some cerebellar and cerebral regions and the severity of the clinical positive manifestations. A relationship between increased FA values of the fronto-occipital and temporolimbic tracts and more severe positive symptoms has been previously observed in nevermedicated patients with schizophrenia 5,10 and in medicated patients with schizophrenia with a history of auditory hallucinations and other positive symptoms. [26][27][28] Overall, these findings suggest that the observed pattern of structural WM changes in patients with schizophrenia is already present before treatment initiation and is associated with the occurrence of positive clinical manifestations.
It is worth noting that MD and FA changes may be secondary to modifications of diffusion either parallel (axial diffusivity) or perpendicular (radial diffusivity) to the principal direction of the tensor. 29 Analysis of directional diffusivities may provide addi-tional information on the underlying mechanisms of WM integrity loss. 29 Myelin breakdown is likely to be associated with increased diffusivity perpendicular to the WM tract, whereas axonal damage has been suggested to reflect diffusivity changes parallel to the primary fiber orientation. 30,31 However, although study of axial and radial diffusivity helps to interpret MD and FA changes, radiopathologic correlations on the basis of investigation of these metrics remain controversial. 32

GM Findings
Drug-naïve patients with schizophrenia had reduced volumes of the temporal, parietal, and occipital cortices and cerebellum. Only a few studies have investigated GM volumes in patients with schizophrenia at the first psychotic episode and who are antipsychotic-naïve, and have shown a pattern of GM volume loss 33 similar to the one we observed.
Reduced volume of the fusiform gyrus (14% of tissue loss) is in line with the results of a neuropathologic study of patients with schizophrenia showing an 18% reduction of neuron attenuation in layers III and V of this region. 34 The fusiform gyrus plays a relevant role in face-processing integrating perception, memory, and emotion. 35 Reduced GM volume of this region may, therefore, be linked to a failure in facial identification and emotion and, thus, may lead to typical impairment of the empathic social interaction known to occur in these patients. 35 Previous structural MR imaging studies in medicated patients with schizophrenia also showed smaller volumes of the posterior cerebellum. 36 Cerebellar damage has been shown to be associated with negative symptoms and mood and behavioral dysregulation. 37 Reduced volume of the precuneus was also found in patients with schizophrenia and was associated with delusions of control. 38 We also detected reduced volume of the superior occipital region, which subtends perceptual functions known to be frequently altered in schizophrenia. In addition, compared with control participants, patients with schizophrenia showed smaller volumes of the middle temporal lobe. This region has been found to be atrophic in a study investigating GM attenuation in high-risk patients in whom psychosis later developed. 39

Study Strengths and Limitations
Having avoided the potential effects played by the chronicity of the disorder on brain MR imaging results and those associated with concomitant treatment, we believe that the presence of multiple structural brain changes in our patients may be conceivably seen as an early phenotypic expression of the processes involved in the pathophysiologic mechanisms of schizophrenia. However, our study was cross-sectional and, therefore, did not allow us to disentangle the role of neurodevelopmental and neurodegenerative abnormalities, which may both result in the observed MR imaging changes. The patients in our study did not undergo comprehensive cognitive testing and were not educationally matched with healthy control participants. However, the inclusion of participants with a Mini-Mental State Examination score Ͼ 24 and IQ Ͼ 75 should have reduced the effect of intellectual functions and educational level on the MR imaging results. In addition, although the exclusion of patients with a lifetime history of Axis I disorders precludes a major effect of cannabis abuse on our find-ings, a possible influence of a sporadic use of cannabis during adolescence cannot be excluded. 1 In patients with recent-onset schizophrenia, a study found that cannabis use before age 17 years was related with increased FA in several brain regions, including the uncinate fasciculus, internal capsule, and frontal WM, compared with patients without a history cannabis use and compared with control participants. 40 In another study, the authors found that compared with cannabis-naïve patients with schizophrenia, patients who used cannabis showed increased FA in the splenium of the corpus callosum. 41 Evidence of an association between increased FA and cannabis use comes from studies of cannabis users without major psychiatric illnesses, showing increased FA 42,43 and reduced MD 42 of several WM regions in these participants. All of these findings are compatible with the ability of the cannabinoid receptor stimulation to increase the oligodendrocyte transcription factor Olig2 as well as the expression of myelin basic protein. 44 Some technical limitations of this study also need to be considered. The diffusion tensor model is currently the framework most commonly used to relate the diffusion signal to the direction of the fibers. However, this model has been shown to be inadequate in voxels containing multiple fiber orientations (eg, "crossing fibers"). 45 Because complex crossing fibers tend to have decreased FA and increased MD, this issue may have influenced our findings. Many approaches have recently been proposed to address this issue, based on high-angular resolution DWI data, 46 and would help to clarify the nature of WM changes in patients with schizophrenia. Head motion can influence MR imaging studies, in particular when dealing with patients with psychiatric disorders. For this reason, structural images (3D T1 and DTI) were carefully checked, and none showed gross head motion artifacts. Furthermore, the DTI preprocessing used in our study includes a movement correction step (see On-line Appendix for further details). Finally, as in previous studies of patients with schizophrenia 47 and in accordance with the exploratory nature of the study design, group differences did not survive multiple comparisons correction, suggesting that structural brain changes in schizophrenia are subtle or that some of the significances may have been overestimated. However, we tried to minimize falsepositive findings by using a relatively conservative statistical threshold in the voxel-based morphometry analysis, and by estimating the standard error of the t statistics by using a permutation approach in the TBSS analysis. In TBSS, we permuted the data 5000 times to obtain large, and consequently relatively conservative, standard errors of the t statistics. Such an approach is usually considered robust enough to allow the use of a canonical statistical threshold of .05. 48