A systematic review of neuroimaging studies of clozapine-resistant schizophrenia

This systematic review aimed to review neuroimaging studies comparing clozapine-resistant schizophrenia patients with clozapine-responding patients, and with first-line antipsychotic responding (FLR) patients. A total of 19 studies including 6 longitudinal studies were identified. Imaging techniques comprised computerized tomography (CT, n = 3), structural magnetic resonance imaging (MRI, n = 7), magnetic resonance spectroscopy (MRS, n = 5), functional MRI (n = 1), single-photon emission computerized tomography (SPECT, n = 3) and diffusion tensor imaging (DTI, n = 1). The most consistent finding was hypo-frontality in the clozapine-resistant group compared with the clozapine-responding group with possible differences in frontal-striatal-basal ganglia circuitry as well as the GABA level between the two treatment-resistant groups. Additional statistically significant findings were reported when comparing clozapine-resistant patients with the FLR group, including lower cortical thickness and brain volume of multiple brain regions as well as lower Glx/Cr level in the dorsolateral prefrontal cortex. Both treatment-resistant groups were found to have extensive differences in neurobiological features in comparison with the FLR group. Overall results suggested treatment-resistant schizophrenia is likely to be a neurobiological distinct type of the illness. Clozapine-resistant and clozapine-responding schizophrenia are likely to have both shared and distinct neurobiological features. However, conclusions from existing studies are limited, and future multi-center collaborative studies are required with a consensus clinical definition of patient samples, multimodal imaging tools, and longitudinal study designs.


INTRODUCTION
The clinical course of schizophrenia shows huge interpatient variability, ranging from complete remission to persistent severe symptoms with significant functional impairment 1,2 .Despite adequate trials of at least 2 antipsychotic medications with sufficient dose and duration, ~15-30% of patients experience persistent symptoms [3][4][5] , and are defined as treatment-resistant schizophrenia (TRS) 6 .At present, clozapine is the only antipsychotic medication that is effective in ameliorating symptoms of TRS 7 with evidence showing clinical efficacy in both short and longterm studies [8][9][10][11] .However, 30-70% of patients with TRS show inadequate response to clozapine 3,12 , and are categorized as ultratreatment-resistant schizophrenia (UTRS) 6 .UTRS patients are associated with poorer clinical and functional outcomes compared to non-treatment-resistant schizophrenia (NTRS) and TRS patients who responded to clozapine 5 .Though delay in clozapine prescription is consistently identified as a factor associated with poor clozapine response 5,13 , multiple barriers to clozapine prescription are suggested including concerns of side effects 14 .Understanding factors associated with clozapine non-response, including neurobiological mechanisms may contribute to the development of targeted interventions to improve outcomes.Of the many studies examining predictors or associated factors contributing to clozapine responses, relatively few consistent effects were reported including younger age, fewer negative symptoms at onset and paranoid schizophrenia subtypes 15 ; consistent biological predictors were lacking 16 .
Earlier studies suggested that a disturbance in dopaminergic transmission is central to schizophrenia 17 , with a growing focus on presynaptic dopaminergic dysfunction 18 .A meta-analysis of 44 studies using positron emission tomography (PET) or singlephoton emission computed tomography (SPECT) reported a significant increase in synthesis and release of striatal dopamine in patients with schizophrenia 19 .Furthermore, the magnitude of dopamine synthesis increase predicted the clinical efficacy of dopamine receptor antagonist antipsychotics, and thus the treatment response 20,21 .In contrast, a cross-sectional study found a lower dopamine synthesis capacity in TRS than NTRS patients 22 .Coupled with the lower affinity of clozapine for dopamine receptors 23 , a different underlying of pathophysiology for TRS was proposed, as well as the possibility of using response to antipsychotic medications to characterize biologically distinct subtypes of illness 24 .
Evidenced by the two meta-analyses of 1 H magnetic resonance spectroscopy ( 1 H-MRS) studies, glutamate hypothesis is a complementary theory of schizophrenia, highlighting the role of upstream N-methyl-D-aspartate (NMDA) receptor hypofunction in causing a downstream cascade of neurotoxicity 25,26 .In more recent cross-sectional studies, TRS patients showed significantly higher glutamate levels in the anterior cingulate cortex (ACC) than the NTRS patients 27 and healthy controls 28 .Moreover, higher ACC glutamate levels were associated with poorer treatment response to non-clozapine antipsychotics 29 .Meta-analyses of structural magnetic resonance imaging (MRI) studies reported subcortical volumetric reduction in schizophrenia patients compared to healthy subjects 30,31 .Specifically, cortical thinning was observed in the dorsolateral prefrontal cortex (DLPFC) of TRS patients compared to NTRS patients 32 .These neuroimaging studies highlighted the possibility of different biological mechanisms in TRS and NTRS and suggest reducing ACC glutamate levels may be a strategy for intervention in TRS.
Over decades, understanding the mechanisms of clozapine in TRS has been challenging due to the heterogeneity of TRS and the complex interaction between receptors and neurotransmitter systems 33 .Despite growing evidence supporting structural brain changes and differential neural mechanisms between NTRS and TRS, the neuroimaging evidence comparing TRS groups, including those who respond to clozapine (CRS) with UTRS is sparse and inconsistent 16 .An improved, systematically derived understanding of the neurobiological substrates of TRS as well as differences between CRS and UTRS would contribute to understanding the mechanisms of clinical efficacy of clozapine in TRS, and could facilitate greater application of existing clozapine treatment as well as facilitate development of novel interventions.
The aim of this review was to address the current research gap by systematically reviewing neuroimaging studies comparing patients with UTRS with other stages of the illness, particularly TRS responding to clozapine to provide insight on potential distinct neural mechanisms among UTRS patients.The results may provide further information on the possible presence of unique neurobiological characteristics of different subtypes of schizophrenia corresponding to characteristics of treatment response.

Literature review
Systematic searches of relevant articles were conducted from the electronic database, including Ovid Medline, Embase, Pubmed, PsychINFO, and Web of Science, from inception to 20 August 2022.An additional search was conducted from 21-30 August 2022 to confirm the inclusion of all relevant studies.Studies that were published in English in a peer-reviewed journal with study samples including schizophrenia, schizoaffective disorder, and schizophreniform disorders according to Diagnostic and Statistical Manual of Mental Health (DSM) or International Classification of Diseases (ICD) criteria, having an operationalized definition of UTRS status with comparison of CRS and patients with other treatment response characteristics using any form of neuroimaging approach were included.All cross-sectional and longitudinal studies were included.Conference abstracts, theses, and editorials were excluded.References from other review articles were examined for relevant studies.The review protocol was registered in the public domain (PROSPERO [International Prospective Register of Systematic Reviews] number: CRD42020203527).
An electronic database search was conducted using the following syntax as search terms: ("ultra-treatment resistant") OR ("ultra-resistant schizophrenia") OR ("non clozapine responder schizophrenia") OR ("clozapine resistant schizophrenia") OR ("treatment refractory schizophrenia") OR ("clozapine nonresponder") AND (magnetic resonance imaging OR MRI OR functional magnetic resonance imaging spectroscopy OR fMRI OR magnetic resonance spectroscopy OR MRS OR voxel-based morphometry OR VBM OR positron emission tomography OR PET or diffusor tensor imaging OR DTI OR single-photon emission computed tomography OR SPECT OR computed tomography OR CT OR Diffusion-weighted magnetic resonance imaging OR DWI).The results of the systematic review are reported based on the PRISMA guidelines (Fig. 1).

Data extraction
After removal of duplicates, one of the reviewers (TP) conducted a first-level screening of the titles and abstracts.References of included studies were screened for eligibility.Eligible studies were selected for full review and data extraction.A second reviewer (SKWC) independently conducted the review process.Inconsistencies were resolved through consensus meetings.
Basic information from the eligible studies was extracted independently by two reviewers, including study characteristics (author, publication year, study design, sample size, demographics of the target population, neuroimaging modality), participant details (number of subjects per group), clinical characteristics, statistical analysis used, and main study outcomes.Due to different definitions of UTRS, details of the inclusion criteria and definition of each patient group were documented.Inconsistencies were resolved through consensus meetings.
To align the different names and definitions of the illness status used in various studies, the following definitions are used in the current review: first-line responders (FLR) are those who show adequate response to non-clozapine antipsychotic medications; treatment-resistant schizophrenia (TRS) are those who failed to show an adequate response to at least 2 previous non-clozapine antipsychotics with at least 6 to 8-week trials; TRS that respond to clozapine (CRS) are TRS patients with symptomatic improvement after receiving adequate clozapine trials; TRS that do not respond to clozapine (UTRS) are TRS who failed to show an optimal response after adequate clozapine trials.Studies were grouped based according to brain region of interest and imaging modalities used.

Study quality
Joanna Briggs Institute (JBI) appraisal tools were used to assess the methodological quality of all included studies.This tool allows consistent and transparent judgment of the quality of parameters including recruitment procedures, sample size, and the degree of appropriateness of statistical analysis.JBI critical appraisal tools for cohort and longitudinal studies were used as appropriate.These tools are comprised of 10 questions that address study design, the methodology, and the statistical analysis used and for each question, the risk of bias assessed using 'Yes', 'No', 'Unclear', and 'Not applicable'.Two raters (T.P and D.M) completed the critical appraisal tool independently and any discrepancies were resolved through discussion.

RESULTS
A total of 308 studies were identified based on search keywords; 19 studies fulfilled the inclusion criteria and were finally included in the current review (Fig. 1).Of the 19 studies, three used computerized tomography (CT), seven used structural magnetic resonance imaging (MRI), five used magnetic resonance spectroscopy (MRS), one reported functional MRI, three used singlephoton emission computerized tomography (SPECT), and one used diffusion tensor imaging (DTI).Six were longitudinal and thirteen were cross-sectional studies (Table 1).All studies reported operational definitions of the UTRS and CRS while nine studies included FLR with clear operational definitions and the total sample size of each study ranging from 22 to 152 (Table 1).Study quality is reported in Supplementary Table 1 (for cross-sectional studies) and 2 (for longitudinal studies).A comprehensive overview of the significant findings is displayed in Fig. 2.

Global brain structural and connectivity
Five studies reported the global structural differences between UTRS patients and patients with other treatment response characteristics (Table 2).An early cross-sectional CT study demonstrated significantly increased global sulcal widening and significantly higher total cortical score in UTRS compared to CRS 34 .However, one CT study with 10 UTRS patients and 26 CRS reported no significant difference in general sulcal widening between the two groups 35 and another CT scan study with 12 UTRS patients and 22 CRS also found no significant differences in ventricular brain ratio between the two groups 36 .Compared with healthy controls (HC) and FLR, patients with both treatment-resistant groups had significantly smaller overall gray matter volume (GM) and all patient groups had smaller whole brain volume and white matter (WM) volume than the healthy controls.However, no significant differences were seen between the UTRS and the CRS 37 .In a longitudinal study of response to clozapine, marginally more cortical thinning of the left medial frontal cortex and the  right middle temporal cortex was seen for the UTRS in comparison with clozapine responders, but overall differences in brain volumes and cortical thickness between patients and healthy controls were reported at either time point 38 .
Only one study reported the white matter microstructure using track-based spatial analysis of diffusion tensor imaging data, and a significantly higher fractional anisotropy (FA) of the right superior longitudinal fasciculus in the UTRS patients compared with the CRS patients was found 39 .However, the significance disappeared after adding head motion as a regressor.Furthermore, CRS patients had overall lower FA than health controls, the FLR, and the UTRS patients, though none of the post-hoc analyses survived corrections for multiple testing 39 .Only one functional MRI study investigated whole brain functional connectivity and reported that UTRS had the weaker connectivity in 3 networks including the cerebella-frontal, cingulo-frontal-temporal, and fronto-parietal when compared with healthy controls but no differences when compared to other patient groups 40 .The connectivity between controls, FLR, and CRS patients was found to be similar.

Structural and functional differences based on brain regions
Frontal lobes.Thirteen studies reported results of analyses of frontal lobes; three used CT, three used MRI, three used SPECT, and four used MRS (Table 3).A larger frontal sulcal widening in UTRS compared with CRS was consistently reported in CT scan studies [34][35][36] .These findings suggest that enlarged frontal sulcal widening could be a unique structural characteristic of UTRS.UTRS showed significantly greater cortical thinning in the left medial frontal cortex than CRS longitudinally 38 , while no differences were seen when investigated cross-sectionally 41 .Another MRI study reported no differences in cortical thickness between UTRS and CRS but a significantly greater mean diffusivity was found in UTRS compared to CRS after controlling for age and gender 42 .However, a greater cortical thinning in the frontal regions was observed in UTRS compared with healthy controls and FLR regardless of age, sex, chlorpromazine equivalent (CPZ) daily dose, and PANSS total scores 41 .In addition, UTRS demonstrated a significantly greater mean diffusivity in the ACC compared to healthy controls after controlling for age and gender 42 .
Three 99 Tc-labeled hexamethyl-propylene-aminoxine (HMPAO) SPECT studies were conducted (Table 3).In the earliest study, neither UTRS nor CRS revealed any prefrontal perfusion changes at follow-up 43 .Shortly after, another study discovered a lower perfusion in the lower and right upper dorsal lateral prefrontal cortex (DLPFC) in UTRS compared to CRS at baseline.Furthermore, no changes in cortical frontal perfusion were observed in the UTRS group, whereas the responder group had a significant reduction in cortical frontal perfusion 44 .The later SPECT study found a significant increase in frontal/caudate perfusion ratio in the clozapine responder group but not in the non-responder group 45 .When compared with HC, a significantly lower perfusion value at the left DLPFC was noted in UTRS at baseline 44 .Furthermore, improvement of digit span-forward was significantly correlated with increase in percentage change in the right frontal/caudate perfusion ratio, whereas a significant relationship between improvement of word fluency and increase in percentage change in both right and left frontal /caudate perfusion ratio was seen.
Three spectroscopy studies found no group differences (clozapine responders or non-responders) in glutamate or Glx    (glutamate and glutamine) levels in the dorsolateral prefrontal cortex (DLPFC) 46 , nor Glx/creatinine (Glx/Cr) level 47 , likewise for the glutathione level in the dorsal anterior cingulate cortex (dACC) 48 (Table 3).One study reported higher Glx levels in the anterior cingulate cortex in the UTRS group compared with clozapine responders (Ochi et al., 2022).Another study reported that UTRS had higher Glx/Cr levels in the DLPFC compared with the FLR 47 .A recent study found a higher gamma-aminobutyric acid (GABA) level in mid-cingulate cortex (MCC) in UTRS compared with CRS after controlling for smoking status, sex, education, GM/ (GM + WM), and age but no differences in Glx levels were found 49 .Furthermore, all four spectroscopy studies reported no significant relationship between the levels of these neuro-metabolites and clinical or cognitive function scores.
Parietal lobes.Two studies (one MRI and one SPECT) examined the parietal lobes (Table 4) without finding a difference in GM and WM volume nor in the perfusion changes between UTRS and CRS 37,45 .When compared with FLR, an extensive reduction of GM volume in the right parietal operculum was observed among UTRS 37 .
Occipital lobes.Three studies (two MRI and one SPECT) investigated the occipital lobes region (Table 4) reporting no significant difference in GM, WM, and CSF between UTRS and CRS 37,41 , and no significant difference in perfusion ratio between the groups 43 .However, UTRS showed a significant GM reduction in the right lateral occipital cortex when compared with healthy controls, whereas the CRS patients showed a significant GM reduction in the lateral occipital cortex when compared with the FLR group 37 .Although no significant differences in cortical thinning in occipital lobes were found between UTRS and CRS, a significantly greater cortical thinning in UTRS was found in occipital gyri when compared with healthy controls, in addition to a more extensive thinning in the left occipital gyri than FLR 41 .
Temporal lobes.Five studies examined temporal lobe regions (Table 4) with only one study reporting significant cortical thinning of the right middle temporal cortex in UTRS compared with CRS 38 .Other studies did not find any difference in cortical thickness 41 , no volumetric differences in lateral ventricle, hippocampus, and amygdala 50 , no GM volume differences of temporal cortex 37 , nor cerebral perfusion differences at the anterior and posterior temporal lobe 43 between the two groups.When compared with FLR and HC, significant cortical thinning was found in UTRS 41 .A bilateral pattern of decreased GM volume was also found in the superior and middle temporal gyri when compared between UTRS and healthy controls 37 .Compared with FLR, a significant reduction of the GM volume in superior, middle, and inferior temporal gyri was seen in the CRS group 37 .
Basal ganglia.Six studies examined the basal ganglia region; three used MRI, one MRS, and two SPECT techniques (Table 4).All three MRI studies reported no significant volumetric differences in the basal ganglia region between UTRS and CRS in both a crosssectional study 51 and longitudinal studies 50,52 .However, UTRS showed a smaller mean striatal volume, globus pallidus, nucleus accumbens, pre-and post-commissural putamen, and pulvinar nucleus volume when compared against FLR 51 .Furthermore, both patient groups were found to have a reduction in volume of putamen and hippocampus compared with healthy controls 50 but no differences were shown between UTRS and HC 51 .
The study examining glutamatergic function reported a significantly higher Glx/Cr in CRS than UTRS in putamen after controlling for CPZ but no other significant differences in metabolites were detected in putamen between UTRS and other comparison groups 47 .Lastly, one 99 Tc-labeled HMPAO SPECT study 44 noted decreased perfusion in the bilateral basal ganglia in UTRS compared to CRS, while the other indicated a decreased perfusion in the left basal ganglia in CRS instead 43 .When compared with HC, UTRS had significantly lower perfusion in the basal ganglia 44 .Furthermore, another 99 Tc-labeled HMPAO SPECT study found a significant increase in right and left frontal/caudate perfusion ratio in the CRS compared with the UTRS 45 .
Thalamus.Four studies examined the thalamus region, two using MRI and two SPECT (Table 4).No significant volumetric differences in the thalamus between UTRS and CRS were observed in a cross-sectional 51 or a longitudinal study 50 .A smaller pulvinar nucleus volume in UTRS compared to FLR and HC and a smaller mean thalamus volume was found between UTRS and HC after controlling for age, sex, total brain volume, education, tobacco use, life history of substance dependence or abuse 51 .In the two 99 Tc-labeled HMPAO SPECT studies, both showed CRS had a significant decrease of perfusion in thalamus 43,44 .
Cerebellum.Only one MRI study examined the cerebellum region (Table 4), reporting no significant GM differences observed between the two treatment-resistant patient groups.A significant reduction of GM volume in the left cerebellum was found comparing the UTRS and the FLR patients as well as UTRS and the healthy controls 37 .

DISCUSSION
Certain neurobiological features of treatment-resistant schizophrenia appear categorically different from treatment-responsive patients and could be considered biomarkers 53 .Response to antipsychotic medications is further suggested as a subtyping strategy for patients with schizophrenia 54 .Despite the plethora of literature on biomarkers in predicting treatment-resistant schizophrenia as well as clozapine response 16 , there were only 19 neuroimaging studies identified in the current review specifically comparing UTRS patients with patients with other characteristics of treatment responsiveness, in particular clozapine response.Among these studies, comparisons of frontal lobe properties were reported most frequently (13 studies) and generated the greatest number of differences between groups.Four out of five studies reported a significant difference between UTRS and CRS, including lowered cortical thickness and widening of sulci, suggesting a general reduction of the frontal lobe volume in UTRS patients in comparison with patients who responded to clozapine.Furthermore, two SPECT studies both reported a significant reduction in perfusion of the frontal region and the frontal/caudate perfusion ratio in UTRS patients.One MRI study reported UTRS had a significantly greater mean diffusivity in ACC than CRS.Studies of other brain regions are much fewer, and all reported no differences in brain volumes.Only one MRS study reported a significantly lower Glx/Cr level in putamen in UTRS compared with CRS, one MRS study found a significantly higher GABA level in mid-cingulate cortex (MCC) in UTRS compared with CRS, and two SPECT study reported lower perfusion and lack of perfusion changes with clozapine treatment in the thalamus among the UTRS patients compared with the CRS patients.These findings suggested that the most pervasive and significant neurobiological differences between the UTRS patients compared with patients who responded to clozapine are likely to be in the frontal region.In fact, this is also reflected by the largely negative findings of global brain volume and cortical thickness comparison between the two treatment-resistant group patients.Furthermore, compared with healthy controls, the UTRS group was found to have lower brain functional connectivity of three networks, all involving the frontal region.But no difference in the CRS patients was found in comparison with controls.These findings align with previous reports on the presence of low prefrontal cortex activities in clozapine-resistant patients 16 and relationship of hypo-frontality  and clozapine treatment in an earlier systematic review 55 .Furthermore, an earlier systematic review identified 5 studies comparing clozapine responders and clozapine non-responders also similarly found involvement of the frontal region 53 .Studies of other regions comparing UTRS and CRS patients are relatively few, and mostly with negative findings apart from basal ganglia and thalamus.One SPECT study reported both reduction in perfusion of the bilateral basal ganglia and thalamus in UTRS patients compared with the clozapine response group and two studies reported a lack of perfusion changes in the UTRS with clozapine initiation.Another SPECT study also suggested significantly lowered frontal/caudate perfusion ratio in the UTRS patients.Furthermore, one MRS study reported significantly higher Glx/Cr in CRS than UTRS in putamen.Previous review also reported the caudate volume and basal ganglia perfusion were related to clinical response to clozapine without specific comparison of UTRS and CRS 53 .Though all the positive findings were reported in only a single study, coupled with the findings of the frontal lobe region, it is possible that frontal-striatal-basal ganglia circuitry function may represent a distinct neurobiological marker of UTRS.However, further exploratory studies are required.
Only one DTI study compared the white matter microstructure of the two treatment-resistant groups.Surprisingly the CRS patients were found to have the lowest FA compared with all other groups (HC, FLP and UTRS).In particular, they had a significantly lower FA of the right superior longitudinal fasciculus compared with the UTRS.Though it was no longer significant after correction for multiple testing and adding the head motion as regressor, the preliminary results of this single study may indicate the possible presence of a unique pattern of white matter microstructure of patients who respond to clozapine.Lower FA indicates lower homogeneity of white matter tractography and poorer white matter integrity.However, a study of Williams syndrome found higher FA of the superior longitudinal fasciculus tract associated with poorer visual-spatial functioning 56 .A counterintuitive result of an early brain imaging study also found brain dysmorphology is related to better symptom improvement with clozapine 57 .These results place more complexity into the relationship between white matter microstructure and the clozapine responses of TRS patients.There were only three MRS studies comparing the two treatment-resistant patient groups and only one study of GABA, that reported significantly higher GABA level in mid-cingulate cortex (MCC) in UTRS compared with CRS.A unique relationship between clozapine and GABA B receptor-mediated inhibitory neurotransmission was reported 58 as well as the binding properties of clozapine with the GABA B receptor 59 .Therefore, it is possible that the GABA level may be a biomarker of clozapineresistant TRS patients.However, replication studies are required.Furthermore, given that the MCC was the only region examined, studies of GABA level in other brain regions particularly in the frontal lobes are needed to further our understanding.Studies of Glx/Cr were largely negative with only one reporting a significantly higher Glx/Cr level in CRS than UTRS in putamen.
There were more significant findings when comparing UTRS with the FLR group including lower cortical thickness and brain volume of multiple brain regions as well as lowered Glx/Cr level in the dorsolateral prefrontal cortex.In fact, both treatment-resistant groups were found to have extensive differences in neurobiological features in comparisons with the FLR group.These suggested that the treatment-resistant patient groups have biologically distinct features compared with the FLR group and are likely to be a subtype of schizophrenia.Within the treatment-resistant patient group, the clozapine response group and clozapineresistance group may share certain neurobiological features.However, a distinct hypo-frontality, abnormalities of frontalstriatal-basal ganglia circuitry as well as the GABA level differences may be neurobiological features differentiating the two groups.One of the strengths of this review study is covering results from multiple imaging modalities whilst focusing on specific regions of the brain to provide a comprehensive coverage of the neurobiological characteristics of UTRS.Secondly, the comparison was focusing specifically on differences between UTRS and CRS as well as FLR and HC and only studies with clearly stated operational definitions of different treatment outcome groups of schizophrenia were included.However, the limited sample size, diverse sample definitions and imaging modalities of existing studies make the conclusions difficult.Furthermore, heterogeneity of the TRS has been reported before, some developed TRS during the first episode and others after multiple relapses 5,60 , thus presence of multiple neurobiological characteristics of TRS and UTRS are possible.Various symptom mixtures of TRS patients might have also contributed to the diversity of results.Therefore, the conclusions with the current literature are far from complete in the pursuit of the understanding of neurobiological nature of different TRS groups and mechanisms of the TRS development.Only five studies adopted a longitudinal design, limiting examination of the effects of clozapine.On the other hand, clinical study of treatment-resistant populations with large sample sizes is challenging.Future multi-center collaborative studies are needed to examine neurobiological markers of clozapine-resistant schizophrenia using the current consensus definition of patient samples 6 , multimodal imaging tools and a longitudinal study design.Studying neurobiological changes because of the neuromodulation interventions of TRS could also be a viable strategy.Moreover, studies with better symptom characteristics of UTRS and focusing on specific symptom dimensions such as hallucination might better inform the neurobiological mechanisms of TRS and UTRS.

CONCLUSIONS
This systematic review of neuroimaging studies comparing clozapine-resistant schizophrenia patients with patients responding to clozapine and patients responding to other antipsychotic medications found 17 studies with variable definitions of patient samples and study methodologies.The most consistent finding was the hypo-frontality of the clozapine-resistant group compared with the clozapine responsive group with a possible difference of frontal-striatal-basal ganglia circuitry as well as the GABA level between the two treatment-resistant patient groups.Extensive neurobiological differences were seen between the two treatment-resistant patient groups and patients responding to other antipsychotics.These suggest the treatment-resistant schizophrenia is likely to be a neurobiological subtype of schizophrenia.Clozapine-resistant and clozapine-response schizophrenia are likely to have some shared neurobiological features, but possible distinct features in the frontal lobe, frontal-striatalbasal ganglia circuitry as well as the GABA level.However, available studies are limited and define the need for multi-center collaborative studies using a consensus definition of patient samples, multimodal imaging tools, and longitudinal study designs.

Table 1 .
Summary of the included study designs and characteristics.

Table 2 .
Global structural differences and functional connectivity.

Table 3 .
Magnetic resonance spectroscopy, structural, functional connectivity differences of frontal lobes and cingulate cortex between UTRS, CRS, FLR, and healthy controls.

Table 4 .
Structural and functional differences between UTRS, CRS, FLR, and healthy controls of brain regions other than the frontal lobes and cingulate cortex.