Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis

SCHRES-07221; No of Pages 8 Schizophrenia Research xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Schizophrenia Research journal homepage: www.elsevier.com/locate/schres

Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis Soo-Hee Choi a,b,c, Sunghyon Kyeong d, Kang Ik K. Cho c,e, Je-Yeon Yun a, Tae Young Lee b,c, Hye Yoon Park a, Sung Nyun Kim a, Jun Soo Kwon a,b,c,e,⁎ a

Department of Neuropsychiatry, Seoul National University Hospital, Republic of Korea Department of Psychiatry, Seoul National University College of Medicine, Republic of Korea Institute of Human Behavioral Medicine, SNU-MRC, Republic of Korea d Severance Biomedical Science Institute, Yonsei University College of Medicine, Republic of Korea e Department of Brain and Cognitive Sciences, Seoul National University College of Natural Sciences, Republic of Korea b c

a r t i c l e

i n f o

Article history: Received 7 September 2016 Received in revised form 9 March 2017 Accepted 10 March 2017 Available online xxxx Keywords: Individuals at clinical high risk for psychosis Network efficiency Structural-functional coupling Disease-specific analysis Neurocognitive function

a b s t r a c t We aimed to separate individuals at clinical high risk for psychosis (CHR) state into subgroups according to neurobiological characteristics using structural and functional network constructs and examine their clinical characteristics. Structural diffusion tensor imaging and resting-state functional magnetic resonance imaging were performed in 61 healthy controls (HC), 57 individuals at CHR and 29 patients with schizophrenia (SZ). The main outcome was a likelihood ratio calculated from measures of structural and functional network efficiencies, coupling strength of structural and functional networks, and a disease-specific data analysis, resulting in the most probable classification of CHR into HC or SZ. The likelihood ratios revealed that 33 individuals at CHR were likely similar to HC (CHR-HC), and the remaining 24 CHR individuals were similar to SZ (CHR-SZ). The CHR subgroups were comparable to each other in demographic characteristics and clinical symptoms. However, the verbal and executive functions of CHR-HC were similar to those of HC, and those of CHR-SZ similar to SZ. Additionally, CHRSZ was more responsive to treatment than CHR-HC during the follow-up period. By combining structural and functional data, we could detect the vulnerable population and provide an active intervention in the early phase of the CHR state. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Individuals at clinical high risk for psychosis (CHR) have been the focus of clinical research for the early detection and prevention of psychotic disorders since the 1990s (Kwon et al., 2012). Longitudinal observation of these individuals, however, showed that 36% of CHR individuals symptomatically remitted and 30% functionally recovered, whereas 30% converted to psychosis within 2 years (Schlosser et al., 2012). This tells us that a heterogeneous group of individuals are in the CHR state between the transient disturbance of mental state in youth and the prodromal stage of psychosis. If we can distinguish the individuals who will have remission or transitory nonpsychotic disorders from those who will undergo overt psychosis or persistent attenuated

⁎ Corresponding author at: Department of Psychiatry, Seoul National University College of Medicine, Department of Brain & Cognitive Sciences, Seoul National University College of Natural Sciences, 101 Daehak-ro, Jongno-gu, Seoul 03080, Republic of Korea. E-mail address: [email protected] (J.S. Kwon).

symptoms, then appropriate intervention in the early stage of non-specific mental distress would be possible, according to the staging model of prodromal prevention (Fusar-Poli et al., 2014). Similar to a risk rating for cardiovascular disease or cancer, Cannon and colleagues (Cannon et al., 2016) developed a “risk calculator” for the personalized prediction of psychosis using clinical, demographic and cognitive measures and demonstrated clinical utility of this calculator (Carrión et al., 2016). In addition to these sets of clinical information, neurobiological measures can improve the individualized approach for detection and treatment in individuals at CHR. Recently, Clementz et al. (2016) reported a possible advantage of neurobiological versus clinical phenomenology for differentiating psychotic disorders. As one of the possible brain-based biomarkers, network models can provide insight into the basic structures and mechanisms that underlie mental illnesses (Park and Friston, 2013; Sporns, 2014). In patients with schizophrenia (SZ), structural and functional brain networks analyses revealed that the connection density among rich club hubs was significantly reduced, suggesting a disruption of global communication in this disease (van den Heuvel et al., 2013).

http://dx.doi.org/10.1016/j.schres.2017.03.028 0920-9964/© 2017 Elsevier B.V. All rights reserved.

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

2

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

We performed a combined structural and functional imaging study using network analysis in accordance with the disconnection model of SZ (Fornito et al., 2012). Structural networks represent anatomical configurations, whereas functional networks represent the interactions among the time series of neuronal activity (Sporns, 2014). A diseasespecific data analysis, termed the Healthy State Model (HSM) (Nicolau et al., 2007; Nicolau et al., 2011), was adopted to measure the error or deviation from the normal state, as well as global and local efficiencies of the network and structural-functional coupling. We aimed to obtain the most probable classification of individual clinical high risk for psychosis into a subgroup that is similar to healthy controls and the classification of the remaining individuals into another subgroup that is similar to patients with schizophrenia, according to structural and functional network constructs. We hypothesized that the resulting subgroups would represent the respective clinical and neurocognitive features of healthy controls or patients with schizophrenia.

directions. One volume was acquired with no diffusion gradient (B0 image). Resting-state functional magnetic resonance imaging was applied over 418 s using a gradient echo-planar imaging sequence. For the other protocols and imaging data processing, see the Supplement. To construct the structural network using fiber tracts, we registered an automated anatomical labeling (AAL) atlas (Tzourio Mazoyer et al., 2002) to the individual high-resolution T1-weighted image using a non-linear transformation matrix obtained from the segmentation steps in the SPM8 package (www.fil.ion.ucl.ac.uk/spm/). The individually fitted AAL map parcellates the brain into 82 cortical regions in the individual space. For the functional network, the whole brain was parcellated into 82 cortical and eight subcortical regions with the AAL atlas. Then, the functional network of each subject (Rij) was computed using Pearson's correlation coefficients between the mean time series of the i-th and j-th regions-of-interest.

2. Materials and methods

For the constructed structural and functional networks, we computed the global and local network efficiencies to examine both global and regional network characteristics (Rubinov and Sporns, 2010). We used non-zero elements to calculate structural network efficiencies and sparsity threshold S (0.08 ≤ S ≤ 0.48) to calculate the functional network efficiency. The structural and functional network efficiency was compared with those of 1000 random networks. For statistical comparisons of the functional network efficiency, the area under the curve (AUC) for all sparsity thresholds was calculated for each global and local efficiency. A structural-functional coupling between the non-zero edges of the structural network and their functional counterparts was obtained using the following procedure. All non-zero entries of the structural network were selected, rescaled to a Gaussian distribution, and correlated with their functional counterparts selected from the Fisher's r-to-z transformed functional network. This produced a single structural-functional coupling value for each brain network (Honey et al., 2009).

2.1. Participants and clinical assessments The study sample consisted of 61 healthy controls (HC, aged 17–35), 57 individuals at CHR (aged 15–33) and 29 patients with SZ (aged 15– 35) who participated in a study conducted at the Seoul Youth Clinic as part of the prospective and longitudinal investigation of CHR and SZ (Kwon et al., 2012). The CHR group was assessed using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) and the Structured Interview of Prodromal Symptoms (SIPS) (Jung et al., 2010), both of which were conducted by psychiatrists. Clinical course of CHR individuals was investigated over six months and one year after the enrollment using the SIPS. Patients with SZ were diagnosed in the first psychotic episode with an onset during the previous year using the SCID-I administered by psychiatrists. Psychotic symptoms were assessed using the Positive and Negative Syndrome Scale (PANSS). The functional level was assessed using the Global Assessment of Functioning (GAF). The HC group was confirmed with the SCID-I Non-Patient Edition. Additional information about our participants is available in the Supplement). All subjects provided written informed consent, and parental consent was obtained for subjects younger than 18 years of age. This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Seoul National University Hospital, Seoul, Korea. 2.2. Neurocognitive tests Measures of attention included the Trail Making Test, Part A (TMTA) and the digit span test. Intelligence quotient was measured using the short forms of the Korean version of the Wechsler Adult Intelligence Scale (K-WAIS; verbal subtests, vocabulary and arithmetic; performance subtests, block design and picture arrangement). A verbal fluency test measured the spontaneous oral generation of words within a given time based on phonemic (phonological fluency) or semantic criteria (semantic fluency). Verbal memory was assessed using the immediate and delayed recall tests of the Korean version of the California Verbal Learning Test (K-CVLT), and visual memory was assessed using the Rey–Osterrieth Complex Figure Test (RCFT) with immediate and delayed scores. The Trail Making Test, Part B (TMT-B) and perseverative errors in the Wisconsin Card Sorting Test (WCST) were used to estimate executive functions. For the TMT and perseverative errors in the WCST, higher scores represented poor performance. 2.3. Imaging data acquisition and network construction Diffusion-weighted images (DWIs) were acquired over 775s with diffusion gradients (b-factor 1000s/mm2) along 64 non-collinear

2.4. Network efficiency and structural-functional coupling

2.5. Healthy state modeling To decompose the disease component from the individual data, we adopted the Disease-Specific Genomic Analysis method and applied it to the structural and functional network data (Nicolau et al., 2007; Nicolau et al., 2011). The HSM was initially introduced in a previous microarray data analysis (Nicolau et al., 2007). Previous studies have shown that this type of linear decomposition analysis can also be used for brain network analysis (Leonardi et al., 2013; Park et al., 2014; Kyeong et al., 2015). Following a previously published method (Kyeong et al., 2015), the upper triangular part of a connection matrix of the network constructs was extracted and vectorized for each subject. The vectorized network data of each subject was stacked in rows and decomposed into normal and disease components through a disease-specific data analysis (Nicolau et al., 2007). Finally, we obtained the magnitude of the disease component in each subject (see Kyeong et al. (2015) for detailed equations). 2.6. Likelihood ratio To estimate the extent of deviation in the structural and functional constructs of individuals at CHR, we determined the likelihood functions for the HC (LHC) and SZ (LSZ) groups by multiplying the probability density functions (PDFs) obtained from the distributions of the global and local network efficiencies of structural and functional networks, the structural-functional coupling, and the magnitude of the disease component from the HSM. These were combined into a likelihood ratio RSZ = LSZ/(LSZ + LHC). We set the decision boundary at a 0.5 likelihood ratio, in which b 0.5 was assigned to the HC group and ≥0.5 was assigned to the SZ group.

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

3

for the local efficiency showed no group differences (F2,144 = 0.100, p = 0.903). The post-hoc analysis indicated that the global efficiency in HC was significantly larger than that in SZ and marginally larger than that in CHR. The above results were maintained when considering the effects of age and sex, except for the local efficiency of the structural network (F2,142 = 2.73, p = 0.069).

2.7. Statistical analysis A one-way analysis of variance (ANOVA) with post-hoc analysis (Tukey's honest significant difference test), analysis of covariance (ANCOVA) and Chi-square tests were used to compare group differences in demographic, clinical, neuropsychological, and network characteristics. Permutation test was used in the comparison of HSM that do not follow a normal distribution. Repeated measures ANOVA was performed to test a group difference in SIPS scores over follow-up periods in CHR subgroups. All statistical analyses were two-tailed, with a significance level of probability set at 0.05.

3.3. Structural-functional coupling Structural-functional coupling strength was significantly different across the three groups (F2,144 = 5.46, p = 0.005). SZ had a significantly reduced structural-functional coupling strength compared with HC (Fig. 1D). The average structural-functional coupling of CHR was between those of HC and SZ; however, the post-hoc analysis did not report any significant differences. The result was unchanged when considering the effects of age and sex.

3. Results 3.1. Demographic and clinical characteristics of participants As shown in Table 1, there were significant differences in age, sex ratio, and education level among the three groups. Because of the clinical characteristics of CHR individuals, the CHR group was younger and had a lower education level than the HC and SZ groups. The SZ group had fewer male participants than the HC and CHR groups. Thus, we additionally performed group comparison using covariates of age and sex to control the potential confounding effects when comparing the three groups. The PANSS and SIPS scores revealed a moderate level of symptoms in the CHR and SZ groups. The duration of illness in the SZ group was shorter than the duration of attenuated psychotic experience in the CHR group because we recruited patients who experienced their first psychotic episode during the previous year. Functional disability was comparable between these two groups. Neurocognitive function of the three groups is reported in the Supplement (Table S1).

3.4. Healthy state modeling The individual brain network was successfully divided into normal and disease components by the HSM. Fig. 2 shows the magnitude of the disease component for each group, which were obtained from the structural and functional networks, as well as from these two networks combined. The result revealed significant group differences in the magnitudes of the disease component of the structural (F2,144 = 23.86, p = 1.12 × 10−9), functional (F2,144 = 173.90, p b 2 × 10−16) and combined structural and functional (F2,144 = 192.10, p b 2 × 10−16) networks. Results of the post-hoc analyses are described in Fig. 2 and the Supplement. The significant group differences were maintained when it was performed with ANCOVA.

3.2. Efficiency of brain network

3.5. Classification of individuals at clinical high risk for psychosis according to the likelihood ratio using structural and functional imaging data

Fig. 1B–C illustrates the global and local efficiencies of the structural and functional networks of participants. In the structural network, there were significant group differences for the global (F2,144 = 4.16, p = 0.018) and local (F2,144 = 3.65, p = 0.028) efficiencies. The post-hoc analysis found that the global efficiency was significantly decreased, and the local efficiency was increased in CHR compared with SZ. In the functional network, the AUC for the global efficiency differed among the three groups (F2,144 = 7.44, p = 0.001), whereas the AUC

Thirty-three individuals at CHR were similar to HC (CHR-HC), and 24 individuals at CHR were similar to SZ (CHR-SZ) (Fig. 3A). There were no group differences in age, sex ratio, or education level between the two subgroups. They showed no group differences in clinical symptoms, duration of attenuated psychotic experience, functional disability, or use of antipsychotics at baseline (Table S2). However, the two subgroups of CHR showed different features in neurocognitive functions; CHR-HC showed a tendency to perform

Table 1 Demographic and clinical characteristics of participants. Measure

HC (N = 61)

Age, year Education, year SIPS, total Positive Negative Disorganized General PANSS, total Positive Negative General DUaP/duration of illness, month Global Assessment of Functioning Chlorpromazine equivalent dose, mg/day

Mean 23.7 14.3 – – – – – – – – – – – –

SD 4.4 1.7

Male sex Right handedness

N 39 59

% 63.9 96.7

F/χ2/t

p

Post-hoc analysis

12.8 5.6 4.6 6.6 5.9 10.9 278.7

13.07 19.15 – – – – – – – – – 6.46 1.68 –

b0.001 b0.001 – – – – – – – – – b0.001 0.100 –

CHR b SZ = HC CHR b SZ = HC – – – – – – – – – – – –

% 27.6 93.1

15.41 3.59

b0.001 0.167

SZ b CHR = HC –

CHR (N = 57)

SZ (N = 29)

Mean 19.8 12.4 34.1 9.3 14.0 4.2 6.7 – – – – 28.8 51.2 125.1a

SD 5.0 2.0

24.4 8.6 106.1

Mean 22.8 13.7 – – – – – 68.0 16.4 16.6 35.0 6.8 47.3 280.9

N 40 49

% 70.2 86.0

N 8 27

SD 3.5 1.7 12.4 3.8 6.3 2.6 4.2

PANSS and Global Assessment of Functioning scores for one patient with SZ were missing. HC, healthy controls; CHR, clinical high risk for psychosis; SZ, schizophrenia; SIPS, Structured Interview for Prodromal Syndromes; PANSS, Positive and Negative Syndrome Scale; DUaP, Mean duration of attenuated psychotic experience before the first assessment in individuals at CHR. a N = 2.

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

4

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

Fig. 1. The outline of network analysis and the results of network constructs. Structural and functional network matrices were constructed using parcellated brain regions with an automated anatomical labeling (AAL) atlas and with deterministic tractography and correlation analysis of blood-oxygen-level-dependent (BOLD) time series, respectively. The global and local network efficiencies were each computed using structural and functional network matrices. A structural-functional coupling between the non-zero edges of the structural network and their functional counterparts was obtained with a correlation analysis (A). In the structural network, the bar graphs represent the mean ± standard deviation of the global and local efficiencies normalized by the corresponding random network (B). The bar graphs show the mean ± standard deviation of the area under the curve for global and local efficiencies of the functional network (C). The bar graph shows the mean ± standard deviation of the functional-structural coupling strength (D). Eglobal, global efficiency; Elocal, local efficiency; CHR, clinical high risk for psychosis; SZ, schizophrenia. † p b 0.06, * p b 0.05, and ** p b 0.005 with the post-hoc analysis.

better than CHR-SZ in most of the neurocognitive tests (Table 2). In particular, there were statistically significant differences in verbal functions; the verbal subtest of intelligence and verbal fluency were comparable between CHR-HC and HC and between CHR-SZ and SZ. Additionally, CHR-HC showed higher scores than SZ, and CHR-SZ showed lower scores than HC. Furthermore, there was a significant difference between the two subgroups in the verbal subtest of intelligence (SZ = CHR-SZ b CHR-HC = HC). As for executive functions measured by the WCST, a pattern similar to that observed for verbal functions was revealed (SZ = CHR-SZ b CHR-HC = HC). For the performance subtest of intelligence and verbal memory, there were no differences between the CHR subgroups (SZ b CHR-SZ = CHR-HC = HC). For the structural network efficiencies, CHR-HC showed significant differences with other groups (global efficiency, CHR-HC b HC = CHRSZ = SZ; local efficiency, SZ = CHR-SZ = HC b CHR-HC). However, the magnitude of the disease component obtained using the structural network revealed no difference between the two CHR subgroups (HC b CHR-HC = CHR-SZ = SZ). The CHR subgroups showed no group differences in their functional network efficiencies. However, the magnitude of the disease component obtained using the functional network was different between the two CHR subgroups. CHR-SZ and SZ had similar levels of the disease component, whereas CHR-HC showed a smaller level of the disease component than SZ. Both CHR-HC and CHR-SZ groups had a larger disease component than HC. There was no group difference in the structural-functional coupling strength in the CHR subgroups. However, the magnitude of the disease component obtained using the combined network revealed a serial pattern in the group

averages (HC b CHR-HC b CHR-SZ = SZ). The results of betweengroup comparison of the neurocognitive functions and network characteristics in Table 2 were maintained when considering the effects of age and sex with ANCOVA. Fifteen CHR-HC and 13 CHR-SZ completed all follow-up evaluations, and there was a significant difference in clinical course between the two subgroups (Fig. 3B). CHR-SZ showed a significant reduction at the sixmonth follow-up compared to CHR-HC, whereas there were no significant differences in SIPS total scores at the baseline and one-year follow-up. There were no significant differences in conversion rate and antipsychotic usage during the follow-up periods (Table S2). 4. Discussion This combined structural and functional study demonstrated that the CHR group could be successfully separated into two subgroups, and each subgroup appeared to be similar to either HC or SZ in terms of their verbal and executive functions, as well as their network properties. The proportion of the CHR-SZ subgroup among the whole CHR group was 42% in the present study. This seems much higher than the 8% transition rate at 2 years, which was reported by 2 of the largest CHR research studies to date (Morrison et al., 2012; Anticevic et al., 2015). However, considering that the 7-year follow-up data revealed that 27% of the CHR individuals converted to psychosis and, 21% underwent persistent attenuated symptoms (Fusar-Poli, 2015), the proportion of the CHR subgroup that was similar to the SZ group would be far more than that predicted by the transition rate.

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

5

Fig. 2. Distribution and Average of the Magnitudes of the Disease Components Obtained by the Healthy State Modeling. CHR, clinical high risk for psychosis; SZ, schizophrenia. ** p b 0.005, and *** p b 0.0005 with the post-hoc analysis.

Impairment in language function is one of the key features of SZ, and the functional neural responses to verbal fluency showed strong potential as a diagnostic marker for SZ (Costafreda et al., 2011). Among the widespread and small-to-moderate degree of impairments in neurocognitive functioning, a deficit in language function, such as a low verbal intelligence, fluency or memory, has been reported as a possible predictor for the progression to psychosis in individuals at CHR (Fusar-Poli et al., 2012; Sumiyoshi et al., 2013; Cannon et al., 2016). Executive function is also one of cognitive domains that has shown the largest effect size of impairment in prior studies of SZ (Aas et al., 2013). As a key cognitive deficit in SZ, impairment of executive function occurs early in the course of the illness (Liu et al., 2011; Bora et al., 2014). We found declining trends in verbal and executive functions in the two CHR subgroups according to the likelihood ratio of SZ (for e.g., SZ = CHR-SZ b CHR-HC = HC). This suggests that the classification of individuals in the CHR group according to their structural and functional properties is consistent with existing findings. Despite a great effort to discover essential neurobiological features that can predict the onset of psychosis (Smieskova et al., 2010; Fusar-Poli et al., 2013), an innovative breakthrough has yet to occur because of an enormous variability in the research findings (Fusar-Poli et al., 2015). As any single lesion cannot explain the pathology of SZ, the

limited information provided by the structural or functional data does not seem to be sufficient to illustrate such disease-specific changes. Thus, there is a need to examine the global characteristics of the entire brain rather than region-of-interest methods. We could discriminate SZ patients from the HC group and classify CHR individuals into subgroups using the structural and functional network properties. Consistent with clinical studies, which have adopted an integrative approach by using features from multiple domains, such as psychotic symptom severity, functional disability and neurocognitive deficits, multimodal neuroimaging data could improve the positive predictability of genuine risk for psychosis in CHR individuals (Cornblatt et al., 2015). The network approach allows an integrative analysis across the structural and functional network domains, and such analyses have shown robust structure-function relationships (Sporns, 2014). To investigate dysfunctional connectivity, the topological properties of brain networks have been widely studied in patients with SZ. Among these properties, efficiency represents a quantitative measure of how well information is transformed within the network (Achard and Bullmore, 2007). The global and local efficiency provide measures of globally integrated and parallel information-processing (e.g., integration), local information processing (e.g., segregation) or network fault tolerance, respectively (Fornito et al., 2012).

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

6

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

Fig. 3. Classification of Participants according to the Likelihood Ratio using Structural and Functional Imaging Data and Clinical Course of Each CHR Subgroup. Thirty-three individuals at CHR who were b0.5 and 24 individuals at CHR who were ≥0.5 in the likelihood ratio. Two of the 61 HC were ≥0.5 and seven of 29 patients with SZ were b0.5 in the likelihood ratio (A). Clinical course of CHR subgroups showed a significant difference in repeated measures analysis of variance (F2,25 = 6.13, p = 0.007; NCHR-HC = 15, NCHR-SZ = 13) during the follow-up period. Post-hoc between-group t-test revealed that total scores on the Structured Interview of Prodromal Symptoms (SIPS) were significantly lower in CHR-SZ than CHR-HC at 6 months follow-up (t = 0.2.48, p = 0.020) (B). CHR, clinical high risk for psychosis; SZ, schizophrenia; CHR-HC, CHR similar to HC; CHR-SZ, CHR similar to SZ. * p b 0.05 with the post-hoc analysis.

The global efficiency of the structural network was reduced in patients with SZ in previous studies (Zalesky et al., 2011; Collin et al., 2014). An earlier study reported in a functional network decreased local and global efficiencies of patients with SZ (Liu et al., 2008; Guye et al., 2010). However, mixed results, including a decreased or comparable local efficiency and a comparable or increased global efficiency in patients with SZ, have been reported in subsequent studies (Fornito et al., 2012). Regional topologies of structural or functional networks have shown more complex and inconsistent results. As for structuralfunctional coupling, SZ patients showed decoupling between structural and functional connectivity in localized networks in one study (Skudlarski et al., 2010) but an increase in the strength of structuralfunctional connectivity coupling in the other study (van den Heuvel et al., 2013). Thus, caution should be taken in interpreting the results regarding the structural or functional properties of the present study. One interesting finding in the structural network of CHR individuals should be noted. The decreased global and increased local efficiencies of the CHR group originated mainly from the CHR-HC group, whereas the CHR-SZ group showed global and local efficiencies that were similar to those of the HC and SZ groups. In other words, the structural network topology of CHR-HC individuals demonstrates high clustering associated with robustness to random error but low parallel information transfer. The structural topology of CHR-HC individuals appears to be pathological as it would be shifted toward regular networks and overconnected to provide the same efficiency of parallel information transfer (Bullmore and Sporns, 2009; Guye et al., 2010). However, this alteration in the structural topology of the CHR-HC group could be a strategic change to cope with mental disturbances. Considering the long period

of attenuated psychotic experience, the CHR group would need a tactic to endure or compensate for the psychic fault, and duration of more than two years would be enough to reflect such structural changes. This can be attained by enhancing the local fault tolerance and sacrificing the global efficiency of parallel information (Fornito et al., 2012). An increased local efficiency of structural network in CHR-HC may imply a relevant compensatory mechanism of neural networks for quasi-psychotic experiences. Critically, there were no significant differences in conversion rate during the follow-up period between the two CHR subgroups. This might be attributed to the effectiveness of an aggressive treatment administered by our clinic. Active interventions, such as psychological, pharmacological, or nutritional aids, might delay or prevent transition to psychosis (Stafford et al., 2013). As stated in our previous paper (Lee et al., 2014), we provided intensive treatment including psychoeducation, cognitive-behavioral therapy and pharmacotherapy to individuals at CHR for quicker abatement of symptoms and more rapid remission from the high risk status. Thus, it was difficult to investigate the natural course of clinical status. Interestingly, however, CHRSZ showed more favorable clinical course than CHR-HC during the first six months, although there were no differences at one-year follow-up. Given that majority of individuals at CHR were taking antipsychotics, rapid reduction of SIPS scores at the initial phase of follow-up would have resulted from the medication effect. This implies that we would weigh the benefit against adverse effect in use of antipsychotics for specific sub-population of CHR, whose characteristics are similar to SZ. An important limitation should be considered in this study. Although previous studies have reported regional abnormalities of

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

7

Table 2 Neurocognitive function and network characteristics of participants according to the results of the likelihood ratio. Measure

Neurocognitive function Attention Trail Making Test, Part A Digit span test Intelligence quotient Verbal subtest Performance subtest Verbal fluency Phonological fluency Semantic fluency Verbal memory K-CVLT, immediate K-CVLT, delayed Visual memory RCFT, immediate RCFT, delayed Executive function Trail Making Test, Part B WCST, perseverative errors Network characteristics Network constructs Eg of the structural network Eloc of the structural network Eg of the functional network Eloc of the functional network Structural-functional coupling strength Magnitude of the disease component In the structural network In the functional network In the combined network

HC (N=61)

CHR

SZ (N=29)

CHR–HC (N=33)

CHR–SZ (N=24)

Mean SD

Mean SD

Mean SD

Mean SD

22.3 12.4 114.5 26.1 26.1

6.6 2.0 12.7 4.2 3.2

23.6 13.0 112.2 26.0 25.6

7.0 2.3 12.5 4.2 3.4

24.9 11.7 105.7 23.1 25.7

10.6 1.9 11.4 4.4 3.5

29.1 10.7 98.4 20.9 22.8

10.6 2.0 13.6 3.4 4.5

45.1 42.6

10.2 9.0

41.8 38.1

11.9 9.2

37.7 34.3

8.0 7.7

34.3 32.1

12.7 13.2

2.6 2.2

12.3 12.8

2.9 2.6

11.7 12.5

2.7 3.0

13.6 13.3

3.0 3.0

12.9 12.5

2.7 3.1

12.9 12.8

52.1 9.0

17.4 5.0

70.4 8.3

24.4 6.4

0.836 3.827 0.384 0.749 0.430

0.063 0.946 0.004 0.016 0.041

0.757 5.017 0.382 0.750 0.412

0.48 2.26 2.45

0.15 0.36 0.40

0.76 3.60 3.86

F

Mean difference on post-hoc analysis A vs B

C vs D

A vs C

B vs D

B vs C

4.54⁎ 7.45⁎⁎⁎ 11.92⁎⁎⁎ 12.75⁎⁎⁎ 6.07⁎⁎

−1.35 −0.61 2.27 0.07 0.55

−4.19 1.02 7.30 2.15 2.91⁎

−2.63 0.69 8.83⁎ 2.98⁎ 0.45

−5.48 2.31⁎⁎⁎ 13.86⁎⁎⁎ 5.07⁎⁎⁎ 2.80⁎

−1.28 1.29 6.56 2.92⁎ −0.10

11.2 8.4

7.82⁎⁎⁎ 11.36⁎⁎⁎

3.27 4.46

3.36 2.11

7.37⁎ 8.31⁎⁎

7.47⁎ 5.96⁎

4.10 3.84

8.7 9.0

3.6 3.5

12.82⁎⁎⁎ 16.98⁎⁎⁎

0.46 0.45

2.94⁎⁎ 3.58⁎⁎⁎

1.04 0.66

3.53⁎⁎⁎ 3.78⁎⁎⁎

0.58 0.21

3.2 2.4

11.7 11.4

3.8 3.8

2.43 2.64

0.65 0.81

1.26 1.41

0.64 0.55

1.25 1.15

−0.01 −0.26

65.1 14.3

22.4 7.8

75.5 13.7

35.5 11.1

7.74⁎⁎⁎ 5.70⁎⁎

−18.28⁎⁎ 0.66

−10.39 −13.03 0.56 −5.25⁎

−5.14 −5.35⁎

5.25 −5.91⁎

0.129 1.567 0.006 0.017 0.061

0.869 3.185 0.382 0.747 0.421

0.027 0.384 0.006 0.017 0.046

0.858 3.536 0.380 0.747 0.395

0.062 1.055 0.006 0.020 0.044

12.546⁎⁎⁎ 16.420⁎⁎⁎ 4.971⁎⁎ 0.219 3.796⁎

0.079⁎⁎⁎ −1.190⁎⁎⁎ 0.002 −0.001 0.018

0.011 −0.351 0.002 −0.001 0.026

−0.101⁎⁎⁎ 1.481⁎⁎⁎ 0.002 0.003 0.017

−0.112⁎⁎⁎ 1.832⁎⁎⁎ 0.001 0.003 −0.009

0.39 0.57 0.61

0.90 3.84 4.26

0.42 0.37 0.46

0.89 4.02 4.41

0.37 0.70 0.64

16.964⁎⁎⁎ −0.281⁎⁎⁎ 0.010 −0.418⁎⁎⁎ −0.127 −0.137 118.998⁎⁎⁎ −1.332⁎⁎⁎ −0.177 −1.578⁎⁎⁎ −0.423⁎ −0.246 137.464⁎⁎⁎ −1.409⁎⁎⁎ −0.149 −1.809⁎⁎⁎ −0.548⁎⁎⁎ −0.399⁎

−0.033 0.642 0.003 0.002 0.009

Neuropsychological data from one individual at CHR were missing. A, HC; B, CHR–HC; C, CHR–SZ; D, SZ. For the Trail Making Test and WCST, higher scores represent poor performance. HC, healthy controls; CHR, clinical high risk for psychosis; CHR–HC, CHR akin to HC; CHR–SZ, CHR akin to SZ; SZ, schizophrenia; K-CVLT, Korean version of the California Verbal Learning Test; RCFT, Rey–Osterrieth Complex Figure Test; WCST, Wisconsin Card Sorting Test; Eg, global efficiency; Eloc, local efficiency. ⁎ p b 0.05. ⁎⁎ p b 0.005. ⁎⁎⁎ p b 0.0005.

network properties in patients with SZ, we did not include these findings in our model. Because altered regional connectivity of parietal, temporal, and occipital cortices, as well as subcortical nuclei, has been found, a diffuse disconnection across nearly all areas is indicative of the pathology observed in SZ (Fornito et al., 2012). Therefore, we focused on the global topology and changes in efficiency for the brain network as a whole. Furthermore, the present results need to be verified in other follow-up studies with a larger sample. To conclude, our results suggest that individuals who have features similar to those of patients with SZ can be identified from among the heterogeneous CHR group using the combined structural and functional network properties. Although they exhibited comparable demographic and clinical characteristics, their verbal and executive functions were impaired compared to the other sub-population of CHR individuals who showed similar features with HC. Given that the CHR-SZ group seems more responsive to treatment than CHR-HC during the followup period, CHR-SZ individuals would need more aggressive intervention (e.g., antipsychotic medication) in the early phase of the CHR state to prevent progression to overt psychosis. The CHR-HC group would also need clinical intervention because they exhibit significant distress or impairment in their functioning as well. However, antidepressants, anxiolytics or psychotherapy would be better than antipsychotics for CHR-HC individuals who have similar neurobiological properties to HC. The present results suggest a way to provide individualized intervention to individuals at CHR.

Disclosure All authors report no financial relationship with any commercial interest. Role of funding source This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (Grant no. 2016R1E1A1A02921618). The funding sources had no further role in the study design, collection, analysis, interpretation of the data, writing of the report, and in the decision to submit this report for publication. The corresponding author had full access to the data in the study and had final responsibility for the decision to submit for publication.

Contributors Soo-Hee Choi helped design and run the study, analyzed and interpreted the data, and wrote the manuscript. Sunghyon Kyeong analyzed the imaging data and undertook statistical analyses. Kang Ik K. Cho conducted imaging acquisition and imaging data processing. Je-Yeon Yun and Tae Young Lee administered clinical interview and managed the clinical data. Hye Yoon Park and Sung Nyun Kim reviewed and extensively commented on the first draft of the manuscript. Jun Soo Kwon undertook the study design and managed the whole procedure of this study. All authors were involved in interpretation of the data analysis. All authors contributed to and have approved the final manuscript.

Conflict of interest The authors have no financial relationship with any commercial interest or other potential conflicts of interest.

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028

8

S.-H. Choi et al. / Schizophrenia Research xxx (2017) xxx–xxx

Acknowledgements

The authors would like to express our gratitude to all the individuals who kindly gave their time to participate in this research. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.schres.2017.03.028. References Aas, M., Dazzan, P., Mondelli, V., Melle, I., Murray, R.M., Pariante, C.M., 2013. A systematic review of cognitive function in first-episode psychosis, including a discussion on childhood trauma, stress, and inflammation. Front. Psych. 4, 182. Achard, S., Bullmore, E., 2007. Efficiency and cost of economical brain functional networks. PLoS Comput. Biol. 3 (2), e17. Anticevic, A., Haut, K., Murray, J.D., Repovs, G., Yang, G.J., Diehl, C., McEwen, S.C., Bearden, C.E., Addington, J., Goodyear, B., Cadenhead, K.S., Mirzakhanian, H., Cornblatt, B.A., Olvet, D., Mathalon, D.H., McGlashan, T.H., Perkins, D.O., Belger, A., Seidman, L.J., Tsuang, M.T., van Erp, T.G., Walker, E.F., Hamann, S., Woods, S.W., Qiu, M., Cannon, T.D., 2015. Association of thalamic dysconnectivity and conversion to psychosis in youth and young adults at elevated clinical risk. JAMA Psychiat. 72 (9), 882–891. Bora, E., Lin, A., Wood, S.J., Yung, A.R., McGorry, P.D., Pantelis, C., 2014. Cognitive deficits in youth with familial and clinical high risk to psychosis: a systematic review and metaanalysis. Acta Psychiatr. Scand. 130 (1), 1–15. Bullmore, E., Sporns, O., 2009. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat. Rev. Neurosci. 10 (3), 186–198. Cannon, T.D., Yu, C., Addington, J., Bearden, C.E., Cadenhead, K.S., Cornblatt, B.A., Heinssen, R., Jeffries, C.D., Mathalon, D.H., McGlashan, T.H., 2016. An individualized risk calculator for research in prodromal psychosis. Am. J. Psychiatry 173 (10), 980–988. Carrión, R.E., Cornblatt, B.A., Burton, C.Z., Tso, I.F., Auther, A.M., Adelsheim, S., Calkins, R., Carter, C.S., Niendam, T., Sale, T.G., 2016. Personalized prediction of psychosis: external validation of the NAPLS-2 psychosis risk calculator with the EDIPPP project. Am. J. Psychiatry 173 (10), 989–996. Clementz, B.A., Sweeney, J.A., Hamm, J.P., Ivleva, E.I., Ethridge, L.E., Pearlson, G.D., Keshavan, M.S., Tamminga, C.A., 2016. Identification of distinct psychosis biotypes using brain-based biomarkers. Am. J. Psychiatry 173 (4), 373–384. Collin, G., Kahn, R.S., de Reus, M.A., Cahn, W., van den Heuvel, M.P., 2014. Impaired rich club connectivity in unaffected siblings of schizophrenia patients. Schizophr. Bull. 40 (2), 438–448. Cornblatt, B.A., Carrión, R.E., Auther, A., McLaughlin, D., Olsen, R.H., John, M., Correll, C.U., 2015. Psychosis prevention: a modified clinical high risk perspective from the Recognition and Prevention (RAP) program. Am. J. Psychiatry 172 (10), 986–994. Costafreda, S.G., Fu, C.H., Picchioni, M., Toulopoulou, T., McDonald, C., Kravariti, E., Walshe, M., Prata, D., Murray, R.M., McGuire, P.K., 2011. Pattern of neural responses to verbal fluency shows diagnostic specificity for schizophrenia and bipolar disorder. BMC Psychiatry 11 (1), 18. Fornito, A., Zalesky, A., Pantelis, C., Bullmore, E.T., 2012. Schizophrenia, neuroimaging and connectomics. NeuroImage 62 (4), 2296–2314. Fusar-Poli, P., 2015. The enduring search for the koplik spots of psychosis. JAMA Psychiat. 72 (9), 863–864. Fusar-Poli, P., Deste, G., Smieskova, R., Barlati, S., Yung, A., Howes, O., Stieglitz, R.-D., Vita, A., McGuire, P., Borgwardt, S., 2012. Cognitive functioning in prodromal psychosis: a meta-analysis. Arch. Gen. Psychiatry 69 (6), 562–571. Fusar-Poli, P., Borgwardt, S., Bechdolf, A., Addington, J., Riecher-Rössler, A., SchultzeLutter, F., Keshavan, M., Wood, S., Ruhrmann, S., Seidman, L.J., Valmaggia, L., Cannon, T., Velthorst, E., De Haan, L., Cornblatt, B., Bonoldi, I., Birchwood, M., McGlashan, T., Carpenter, W., McGorry, P., Klosterkötter, J., McGuire, P., Yung, A., 2013. The psychosis high-risk state: a comprehensive state-of-the-art review. JAMA Psychiat. 70 (1), 107–120. Fusar-Poli, P., Yung, A.R., McGorry, P., Van Os, J., 2014. Lessons learned from the psychosis high-risk state: towards a general staging model of prodromal intervention. Psychol. Med. 44 (01), 17–24. Fusar-Poli, P., Cappucciati, M., Borgwardt, S., Woods, S.W., Addington, J., Nelson, B., Nieman, D.H., Stahl, D.R., Rutigliano, G., Riecher-Rössler, A., Simon, A.E., Mizuno, M., Lee, T.Y., Kwon, J.S., Lam, M.M., Perez, J., Keri, S., Amminger, P., Metzler, S., Kawohl, W., Rössler, W., Lee, J., Labad, J., Ziermans, T., An, S.K., Liu, C.C., Woodberry, K.A., Braham, A., Corcoran, C., McGorry, P., Yung, A.R., McGuire, P.K., 2015. Heterogeneity of psychosis risk within individuals at clinical high risk: a meta-analytical stratification. JAMA Psychiat. 1–8.

Guye, M., Bettus, G., Bartolomei, F., Cozzone, P.J., 2010. Graph theoretical analysis of structural and functional connectivity MRI in normal and pathological brain networks. MAGMA 23 (5–6), 409–421. van den Heuvel, M.P., Sporns, O., Collin, G., Scheewe, T., Mandl, R.C., Cahn, W., Goñi, J., Pol, H.E., Kahn, R.S., 2013. Abnormal rich club organization and functional brain dynamics in schizophrenia. JAMA Psychiat. 70 (8), 783–792. Honey, C.J., Sporns, O., Cammoun, L., Gigandet, X., Thiran, J.P., Meuli, R., Hagmann, P., 2009. Predicting human resting-state functional connectivity from structural connectivity. Proc. Natl. Acad. Sci. U. S. A. 106 (6), 2035–2040. Jung, M.H., Jang, J.H., Kang, D.H., Choi, J.S., Shin, N.Y., Kim, H.S., An, S.K., Shin, M.S., Kwon, J.S., 2010. The reliability and validity of the korean version of the structured interview for prodromal syndrome. Psychiatry Investig. 7 (4), 257–263. Kwon, J.S., Byun, M.S., Lee, T.Y., An, S.K., 2012. Early intervention in psychosis: insights from Korea. Asian J. Psychiatr. 5 (1), 98–105. Kyeong, S., Park, S., Cheon, K.A., Kim, J.J., Song, D.H., Kim, E., 2015. A new approach to investigate the association between brain functional connectivity and disease characteristics of attention-deficit/hyperactivity disorder: topological neuroimaging data analysis. PLoS One 10 (9), e0137296. Lee, T.Y., Kim, S.N., Chon, M.-W., Kwon, J.S., 2014. Effects of the functioning and antipsychotic use on clinical high risk for psychosis: a response to Yung et al. Schizophr. Res. 159 (1), 254–255. Leonardi, N., Richiardi, J., Gschwind, M., Simioni, S., Annoni, J.M., Schluep, M., Vuilleumier, P., Van De Ville, D., 2013. Principal components of functional connectivity: a new approach to study dynamic brain connectivity during rest. NeuroImage 83, 937–950. Liu, Y., Liang, M., Zhou, Y., He, Y., Hao, Y., Song, M., Yu, C., Liu, H., Liu, Z., Jiang, T., 2008. Disrupted small-world networks in schizophrenia. Brain 131 (4), 945–961. Liu, K.C., Chan, R.C., Chan, K.K., Tang, J.Y., Chiu, C.P., Lam, M.M., Chan, S.K., Wong, G.H., Hui, C.L., Chen, E.Y., 2011. Executive function in first-episode schizophrenia: a three-year longitudinal study of an ecologically valid test. Schizophr. Res. 126 (1), 87–92. Morrison, A.P., French, P., Stewart, S.L., Birchwood, M., Fowler, D., Gumley, A.I., Jones, P.B., Bentall, R.P., Lewis, S.W., Murray, G.K., Patterson, P., Brunet, K., Conroy, J., Parker, S., Reilly, T., Byrne, R., Davies, L.M., Dunn, G., 2012. Early detection and intervention evaluation for people at risk of psychosis: multisite randomised controlled trial. BMJ 344. Nicolau, M., Tibshirani, R., Børresen-Dale, A.L., Jeffrey, S.S., 2007. Disease-specific genomic analysis: identifying the signature of pathologic biology. Bioinformatics 23 (8), 957–965. Nicolau, M., Levine, A.J., Carlsson, G., 2011. Topology based data analysis identifies a subgroup of breast cancers with a unique mutational profile and excellent survival. Proc. Natl. Acad. Sci. U. S. A. 108 (17), 7265–7270. Park, H.-J., Friston, K., 2013. Structural and functional brain networks: from connections to cognition. Science 342 (6158), 1238411. Park, B., Kim, D.S., Park, H.J., 2014. Graph independent component analysis reveals repertoires of intrinsic network components in the human brain. PLoS One 9 (1), e82873. Rubinov, M., Sporns, O., 2010. Complex network measures of brain connectivity: uses and interpretations. NeuroImage 52 (3), 1059–1069. Schlosser, D.A., Jacobson, S., Chen, Q., Sugar, C.A., Niendam, T.A., Li, G., Bearden, C.E., Cannon, T.D., 2012. Recovery from an at-risk state: clinical and functional outcomes of putatively prodromal youth who do not develop psychosis. Schizophr. Bull. 38 (6), 1225–1233. Skudlarski, P., Jagannathan, K., Anderson, K., Stevens, M.C., Calhoun, V.D., Skudlarska, B.A., Pearlson, G., 2010. Brain connectivity is not only lower but different in schizophrenia: a combined anatomical and functional approach. Biol. Psychiatry 68 (1), 61–69. Smieskova, R., Fusar-Poli, P., Allen, P., Bendfeldt, K., Stieglitz, R.D., Drewe, J., Radue, E.W., McGuire, P.K., Riecher-Rössler, A., Borgwardt, S.J., 2010. Neuroimaging predictors of transition to psychosis–a systematic review and meta-analysis. Neurosci. Biobehav. Rev. 34 (8), 1207–1222. Sporns, O., 2014. Contributions and challenges for network models in cognitive neuroscience. Nat. Neurosci. 17 (5), 652–660. Stafford, M.R., Jackson, H., Mayo-Wilson, E., Morrison, A.P., Kendall, T., 2013. Early interventions to prevent psychosis: systematic review and meta-analysis. BMJ 346, f185. Sumiyoshi, T., Miyanishi, T., Seo, T., Higuchi, Y., 2013. Electrophysiological and neuropsychological predictors of conversion to schizophrenia in at-risk subjects. Front. Behav. Neurosci. 7, 148. Tzourio Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., Mazoyer, B., Joliot, M., 2002. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage 15 (1), 273–289. Zalesky, A., Fornito, A., Seal, M.L., Cocchi, L., Westin, C.F., Bullmore, E.T., Egan, G.F., Pantelis, C., 2011. Disrupted axonal fiber connectivity in schizophrenia. Biol. Psychiatry 69 (1), 80–89.

Please cite this article as: Choi, S.-H., et al., Brain network characteristics separating individuals at clinical high risk for psychosis into normality or psychosis, Schizophr. Res. (2017), http://dx.doi.org/10.1016/j.schres.2017.03.028