Introduction
Schizophrenia is a heterogeneous disease comprising positive symptoms, such as delusions and hallucinations; negative symptoms, such as avolition and social withdrawal; cognitive impairment; and mood dysregulation. While antipsychotic medication is the mainstay of treatment for this condition, it is ineffective for approximately 30% of patients. Treatment-resistant schizophrenia (TRS) is defined as the failure to respond to at least two antipsychotic drug trials administered at an adequate dose for an appropriate period of time.Reference Howes, McCutcheon and Agid 1 Resistance to treatment may cause significant personal, family, and social difficulties, resulting in increased hospitalization rates, longer hospital stays, and the significant consumption of other resources.Reference Kennedy, Altar, Taylor, Degtiar and Hornberger 2 , Reference Revicki 3 These manifold consequences underscore the need for effective therapeutic intervention.
Several lines of evidence indicate that TRS may be a distinct subtype of schizophrenia with different neurochemical abnormalities. This hypothesis is inferred by the observations that responders to treatment present more pronounced dopaminergic abnormalities, while nonresponders present dysfunction in the glutamatergic system.Reference Gillespie, Samanaite, Mill, Egerton and MacCabe 4 , Reference Demjaha, Egerton and Murray 5 Also, clozapine, an atypical antipsychotic with weak dopamine antagonism,Reference Yilmaz, Zai and Hwang 6 remains the reference standard treatment for TRS.Reference Van Sant and Buckley 7 , Reference Elkis and Meltzer 8
White matter (WM) studiesReference Kelly, Jahanshad and Zalesky 9 –Reference Cheung, Cheung and McAlonan 12 have reported abnormalities supporting the hypothesis of dysconnectivity in schizophrenia. The effect of antipsychotics on myelin is not well established. Bartzokis et alReference Bartzokis, Lu and Stewart 13 suggested that the choice of antipsychotics may impact the myelination of posterior intracortical circuits in adults with schizophrenia. Other research suggests that antipsychotics increase intracortical myelin early in the course of their administration.Reference Tishler, Bartzokis and Lu 14 , Reference Reis Marques, Taylor and Chaddock 15 Garver et alReference Garver, Holcomb and Christensen 16 demonstrated that an antipsychotic-induced cascade might partially restore myelin integrity and concomitant functional connectivity in TRS patients.
Understanding of the myelinated fibers connecting cortical and subcortical grey matter (GM) has advanced with the advent of magnetic resonance diffusion tensor imaging (DTI).Reference Patel, Mahon, Wellington, Zhang, Chaplin and Szeszko 17 This can detect changes in WM fiber systems even in the absence of macrostructural changes.Reference Assaf and Pasternak 18 , Reference Heng, Song and Sim 19 One comparison of patients with TRS to healthy controls found the former to have reduced fractional anisotropy (FA) and increased radial diffusivity (RD) in the genu, body, and splenium of the corpus callosum (CC), the right posterior limb of the internal capsule, the right external capsule, and the right temporal inferior longitudinal fasciculus.Reference Holleran, Ahmed and Anderson-Schmidt 20
However, few neuroimaging studies have considered differences between individuals with TRS and those with non-TRS. It thus remains unclear whether the neurobiological profiles of these two populations are distinct. In the present study, we compared microstructural abnormalities in the fiber tracts between TRS and non-TRS. We hypothesized that reduction in FA values in TRS would be greater than in non-TRS.
Materials and methods
Subjects
This study recruited a cross-sectional sample of multiepisode patients with TRS (n = 34) and non-TRS (n = 51) from the outpatient Schizophrenia Program (PROESQ) of the Universidade Federal de São Paulo (São Paulo, Brazil) between 2011 and 2015. Inclusion criteria were as follows: (1) a research interview confirming the diagnosis of schizophrenia according to the DSM-IV (American Psychiatric Association, 1994), (2) age between 16 and 60 years, (3) no neurological disease, (4) no severe intellectual disability, and (5) no comorbidity with other axis I disorders. The Research Ethics Committee approved the study of the Federal University of São Paulo (UNIFESP) (protocol numbers 0661/11 and 1737/06). All participants or their caregivers provided written informed consent prior to inclusion.
Structured diagnostic and symptom evaluation
The diagnosis was confirmed using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I).Reference First, Gibbon, Spitzer, Williams and Benjamin 21 The Positive and Negative Syndrome Scale (PANSS) was administered to all patients to evaluate psychopathology. Sociodemographic information and medical histories were collected from the patients and confirmed by their relatives or caregivers.
Outpatient sample: clinical assessment and resistance to treatment
The outpatients were assessed by four independent trained raters who were regular members of the medical staff and participated in routine clinical decisions. Response to previous antipsychotics was determined by a consensus of at least two experienced clinicians in routine meetings. Diagnosis of TRS followed the criteria established by the International Psychopharmacology Algorithm Project (IPAP): (1) persistence of symptoms after at least two antipsychotics administered for at least 4 weeks, each at doses equivalent to chlorpromazine 400 mg/d or risperidone 5 mg/d; (2) persistence of at least two symptoms of moderate or greater severity; or at least one symptom with at least a severe rating among the following PANSS items: delusions (P1), conceptual disorganization (P2), hallucinatory behavior (P3), or suspiciousness (P6). Additional information on antipsychotic treatment was collected from the patients’ medical records and other staff members. Information about treatment adherence was obtained from caregivers, staff members, and medical records.
MRI data acquisition
All brain scans were obtained at the Department of Diagnostic Radiology (UNIFESP) on a 1.5 T Siemens scanner (Magnetom Sonata AG, Medical Solutions, Erlangen, Germany) with an eight-channel head coil (MAGNETON Sonata). For each patient, diffusion-weighted images of the whole brain in the axial orientation were obtained according to the following parameters: FOV = 256 mm, 50 slices, 128 × 128 matrix, TR = 7000 ms, TE = 85 ms, b-value = 1000 s/mm2, slice thickness = 3.0 mm, and 12 noncollinear diffusion directions.
Image analysis
Data were preprocessed following the basic pipeline available on FMRIB’s Software Library (FSL) (www.fmrib.ox.ac.uk/fsl).Reference Smith, Jenkinson and Woolrich 22 All images were corrected for eddy currents using FMRIB’s Diffusion Toolbox (FDT). Skull stripping was then performed by Brain Extraction Toolbox (BET). Diffusion tensors were subsequently fitted for each voxel (using the DTIFIT tool) to calculate the FA, RD, axial diffusivity (AD), and mean diffusivity (MD) maps.
Whole-brain voxel-wise cross-subject analyses were performed using Tract-Based Spatial Statistics (TBSS).Reference Smith, Jenkinson and Johansen-Berg 23 All FA maps were nonlinearly registered to a Montreal Neurological Institute (MNI) standard space target image and were used to create an average skeleton, and the FA maps of each subject were then projected onto a standard mean FA skeleton at a threshold of 0.2. The same procedure was repeated for the RD, AD, and MD maps.
We then performed the automated extraction of the mean FA value from all regions. This extraction was based on the Johns Hopkins University atlas of WM tractographyReference Wakana, Caprihan and Panzenboeck 24 , Reference Mori and Crain 25 and the atlas published by the International Consortium of Brain Mapping (ICBM) DTI-81.Reference Mori, Oishi and Jiang 26
Statistical analyses
A threshold-free cluster enhancement (TFCE) approach was used for statistical analyses to identify the main effect of the groups (TRS vs non-TRS).
We performed a TFCE by applying 10 000 permutations on unsmoothed statistical maps using the Randomize tool.Reference Winkler, Ridgway, Webster, Smith and Nichols 27 The level of significance corrected for the family-wise error (FWE) rate was set to P < .05. To identify potential confounding effects, we included the factors of age and sex in the GLM matrix. The voxel-wise statistical significance level corrected for the FWE was set to P < .05. The same procedure was used to assess the RD, AD, and MD maps. The tool cluster of the FSLv6.0Reference Jenkinson, Beckmann, Behrens, Woolrich and Smith 28 was used to report cluster information.
To identify differences in demographic and clinical characteristics, we used SPSS 23.0 to perform one-way ANOVA or chi-square tests (Table 1). The level of significance was set to P < .05.
Table 1. Sociodemographic and Clinical Characteristics
Abbreviations: PANSS, Positive and Negative Syndrome Scale; SD, standard deviation.
A stepwise, confirmatory multivariate analysis of FA values extracted from regions of interest was performed to investigate the effect of the relationship between antipsychotic dose and the regions of interest found in the exploratory analysis with SPSS 23.0. Sex, age, duration of illness, antipsychotic dose (chlorpromazine dose equivalent), and the PANSS positive and negative scores were used as independent variables. The FA scores of the splenium of the CC was used as the dependent variable.
Results
The demographic characteristics of the participants are shown in Table 1. Patients with TRS and non-TRS did not differ in age, sex, duration of illness, or the age of onset. However, individuals with TRS were more symptomatic across all symptom domains. All TRS patients were receiving clozapine. Non-TRS individuals received the following antipsychotics: olanzapine (59.2%), risperidone (16.3%), aripiprazole (8.2%), quetiapine (8.2%), chlorpromazine (2.0%), haloperidol (2.0%), and long-acting risperidone (2.0%).
Whole-brain analysis in TBSS
The whole-brain analysis in TBSS showed that in TRS patients, there were a significant FA reduction in the splenium of the CC than did non-TRS (P < .05, FWE corrected; cluster size, 8825; location of maximum z score [x, y, z], −19, −34, 32; Figure 1). There was no increased FA in TRS compared with non-TRS. No differences were found between the two populations in other DTI measures (RD, AD, MD).
Figure 1. Cluster of reduced FA in the splenium of CC in TRS as compared with non-TRS (x = −19, y = −34, z = −32). Significant cluster (P < .05, corrected by TFCE) highlighted in red; yellow was shown on the mean FA image.
Confirmatory analysis
We found no relationship between antipsychotic dose and the FA scores of splenium of the CC.
Discussion
Our results partially confirm our major hypothesis: TRS patients present a significantly lower FA in the splenium of the CC relative to non-TRS patients. However, no differences in RD¸ AD, or MD maps were found.
Reduced FA in the splenium of the CC seems to be the most consistent finding in the schizophrenia literature to date.Reference Aydin, Ucok and Guler 29 –Reference Rotarska-Jagiela, Schönmeyer, Oertel, Haenschel, Vogeley and Linden 31 Decreases in FA could be attributed to lower myelination, neuronal fiber density, or directional coherence.Reference Basser 32 Few neuroimaging studies have compared TRS to non-TRS. However, as the main interhemispheric commissure of WM, aberrations in the CC causing abnormal interhemispheric connectivity have been implicated in the neuropathophysiology of schizophrenia and could be central in TRS.Reference Rotarska-Jagiela, Schönmeyer, Oertel, Haenschel, Vogeley and Linden 31 Anomalies have been reported in the CCs of patients with TRS relative to healthy controls.Reference Holleran, Ahmed and Anderson-Schmidt 20 , Reference Sun, Maller, Daskalakis and Furtado 33 However, while a meta-analysis of 15 studies found that patients with TRS presented significant FA reductions in their CCs relative to healthy controls, this finding was inconsistent across the investigations considered.Reference Crocker and Tibbo 34 While a DTI whole-brain analysis has also revealed decreased FA values in the CCs of patients with TRS,Reference Holleran, Ahmed and Anderson-Schmidt 20 , Reference Mitelman, Torosjan and Newmark 35 , Reference Mitelman, Nikiforova and Canfield 36 none were found in other studies.Reference Chen, Ke and Zhuo 37
RD increases in the CC have also been observed in patients with TRS relative to healthy controls, suggesting that patients with schizophrenia exhibit deficits in microstructural organization.Reference Holleran, Ahmed and Anderson-Schmidt 20 Kochunov and collaborators found a pattern of regional WM deficit in schizophrenia that was significantly associated with resistance to treatment, suggesting that the extent of regional WM vulnerability can already be observed in schizophrenia since the initial diagnosis and treatment. Also, these findings may be a marker of resistance to treatment with antipsychotic medications currently available; however, this hypothesis has not been tested longidutinally.Reference Kochunov, Huang and Chen 38
In agreement with our findings, Mitelman et al found a decrease in the FA of the splenium of the CC of patients with TRS relative to those of patients with non-TRS;Reference Mitelman, Torosjan and Newmark 35 a follow-up study revealed that TRS patients showed a more pronounced decline in the size, but a less pronounced decline in the anisotropy, of the CC relative to their counterparts with non-TRS, suggesting that changes in the CC of the former occur earlier in the course of the disease—closer to the first psychotic episode—but stabilize in the chronic phases of the condition.Reference Mitelman, Nikiforova and Canfield 36 These findings support the role of a lower FA in the splenium of the CC as a biomarker of TRS.
A meta-analysis of voxel-based DTI studies has also reported reduced FA in the CCs of patients with schizophrenia relative to healthy controls.Reference Ellison-Wright and Bullmore 39 Similarly, a meta-analysis found GM and WM abnormalities in the CCs of patients with schizophrenia.Reference Bora, Fornito and Radua 40 The ENIGMA DTI study of microstructural white matter changes also found extensive CC changes and FA reductions in 20 of the 25 regions of interest, involving all major WM fasciculi,Reference Kelly, Jahanshad and Zalesky 9 supporting the hypothesis of structural dysconnectivity in schizophrenia.
Our analysis did not detect increases in clozapine-naïve RD, AD, or MD in the CCs of patients with TRS relative to those with non-TRS; hence, our study failed to show differences in microstructural organization between the two populations. However, no DTI measurement corresponds to one specific property; all diffusion parameters are considered to indicate the status of several tissue properties, such as myelinization, axonal orientation, and axonal density.Reference Alexander, Lee, Lazar and Field 41 , Reference Jones, Knösche and Turner 42 Indeed, research has found that changes in white matter microstructure pathologies may cause unpredictable changes in AD and RD that do not correspond to the tissue organization; hence, these diffusion indices may not always be reliable.Reference Wheeler-Kingshott and Cercignani 43
This study benefits from having examined a larger, more homogeneous sample than have similar studies. Our results thus feature relatively robust validity. While our findings may have been confounded by the exposure of patients to psychotropic medications, the observed changes were not correlated to the dose of antipsychotic medication received. We used no objective adherence measures to ensure compliance in the trials before the TRS diagnosis. We used the 12-direction protocol in a 1.5 T field, although DTI protocols with more diffusion directions may be optimal. Seo et alReference Seo 44 demonstrated that the FA values in large white matter bundles, such as the CC, were not affected by the number of gradient directions, acquisitions, and field strengths. The present study was further limited by not having included healthy controls for comparison in the study. As the current report is cross-sectional in nature, longitudinal studies are required to answer these questions and assess changes in TRS patients over time.
Conclusion
In conclusion, the present study detected a localized reduction in the FA of the CCs of a group of patients with TRS. The diminished FA of the CC may thus feature potential as a biomarker for TRS.
Financial Support
This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brazil (CAPES)—Finance Code 001, and Fundação de Amparo à Pesquisa do Estado de São Paulo—Brazil (FAPESP) S.I.B. 14/07280-1 and 10/08968-6.
Disclosures
Dr. Rodrigo Bressan reports grants, personal fees, and non-financial support from Janssen, personal fees from Lundbeck, others from Roche, personal fees from Ache, and personal fees from Novartis, outside the submitted work. Idaiane Batista de Assunção-Leme, André Zugman, Luciana Monteiro de Moura, João Ricardo Sato, Cinthia Higuchi, Bruno Bertolucci Ortiz, Cristiano Noto, Vanessa Kiyomi Ota, Sintia Iole Belangero, Nicolas A. Crossley, Andrea Jackowski, and Ary Gadelha declare no conflicts of interest inherent to this article.
