A unified model of shared brain structural alterations in patients with different mental disorders who experience own‐thought auditory verbal hallucinations—A pilot study

Abstract Objective To explore shared brain structural alterations in patients diagnosed with mental disorders who experience own‐thought auditory verbal hallucinations (OTAVHs). Methods A cohort of 143 first‐diagnosis, nonmedicated patients with OTAVHs was enrolled: 25 with schizophrenia (FUSCH‐OTAVH), 20 with major depression disorder (FUMDD‐OTAVH), 28 with bipolar disorder (FUBD‐OTAVH), 22 patients with posttraumatic stress disorder (FUPTSD‐OTAVH), 21 with anxiety disorder (FUAD‐OTAVH), and 27 with borderline personality disorder (FUBPD‐OTAVH); 25 healthy controls (HCs) participated. The Auditory Hallucinations Rating Scale (AHRS), multiple psychometric scales, voxel‐based morphometry (VBM), tract‐based spatial statistics (TBSS), and multiple regression were used. Results Compared with HCs, patients had increased occipital cortex, dorsal prefrontal cortex (PFC), and striatum gray matter volumes (GMVs), a reduced insular cortex (IC) GMV, and an impaired frontooccipital fasciculus. The following differences were found versus HCs: FUSCH‐OTAVH, reduced PFC and occipital GMVs, increased striatum and thalamus GMVs, impaired arcuate fasciculus, u‐shaped bundle, optic tract, and upper longitudinal fasciculus (LF); FUMDD‐OTAVH, increased posterior frontotemporal junction and hippocampus GMVs; FUMN‐OTAVH, increased posterior frontotemporal junction and parietal cortex GMVs, reduced hippocampus GMV, impaired upper LF; FUPTSD‐OTAVH, increased temporal, hippocampus, and nucleus accumbens GMVs; FUBPD‐OTAVH, increased frontotemporal junction and hippocampus GMVs, impaired upper/lower LF; and FUAD‐OTAVH, increased frontal and temporal cortex, hippocampus GMVs. Conclusions The present findings provide evidence consistent with a bottom‐up and top‐down reciprocal action dysfunction hypothesis of AVHs and with the dopamine hypothesis of AVHs. We observed specific features related to OTAVHs in patients with different mental disorders. The findings, though complex, provide clues for further studies of specific mental disorders.

Among the aforementioned AVH types, OTAVHs appear to be the most prevalent across different mental disorders (McCarthy-Jones et al., 2014;Upthegrove et al., 2016). Based on clinical phenomenon cluster analysis findings, McCarthy-Jones and colleagues have characterized OTAVHs as consisting of the following properties: hallucinated verbiage not addressing the person hearing them directly, first-person voice syntax, experience similar to a memory and/or one's own inner voice or thoughts (McCarthy-Jones et al., 2014).

| Participants
This study was approved by the Ethics Committee at Tianjin Mental Health Centre, whose patient database was used as our source of study enrollees. Written informed consent was obtained from all participants and their legal guardians (Chinese requirement) prior to data acquisition.
The inclusion criteria for the mental disorder diagnosis groups were as follows: OTAVHs (first-person voices, not addressing the person similar to a memory, and possibly one's own voice/thoughts (McCarthy-Jones et al., 2014)) during MRI; mental disorder diagnosed by two senior psychiatrists according to the DSM-IV (Tong & Phillips, 2010;Tong, Phillips, & Conner, 2016) Applying the aforementioned criteria, we enrolled 25 FUSCH-OTAVH, 20 FUMDD-OTAVH, 28 FUBD-OTAVH, 22 FUPTSD-OTAVH, 21 FUAD-OTAVH, and 27 FUBPD-OTAVH patients as well as 28 HCs. We identified three healthy individuals with OTAVHs, but they were not included due to being too few in number.

| Symptom assessment
OTAVH symptoms were assessed with the Auditory Hallucinations Rating Scale (AHRS) (Wahab et al., 2015). The MATRICS Consensus Cognitive Battery (MCCB) (Lystad et al., 2016) was used to assess cognitive ability. The Global Assessment of Functioning scale (GAF) (American Psychiatric Association, 1994) was used to assess global function. The Positive and Negative Syndrome Scale (PANSS) (Kay, Fiszbein, & Opler, 1987) was used to assess schizophrenia symptoms.
The Hamilton Rating Scales for Depression (Leucht et al., 2013) and Anxiety (Hamilton, 1959) (HAMD and HAMA) were used to assess clinical depression and anxiety disorder symptoms, respectively. The Young Mania Rating Scale (YMRS) (Young, Biggs, Ziegler, & Meyer, 1978) was used to assess the manic symptoms. The Clinicianadministered PTSD Scale (CAPS) (Blake et al., 1995) was applied to assess PTSD symptoms. The sociodemographic characteristics and clinical symptoms of each group are summarized in Table 1.

| MRI acquisition and image processing for VBM
Imaging data were obtained in a 3.0-T Discovery MR750 scanner Gradient nonlinearity-related image distortion was corrected with GradWarp, and intensity inhomogeneity was corrected with the N3 program. In preparation for automated brain-wide VBM, 3D-FSPGR images were subjected to bias correction, spatial normalization, tissue segmentation (gray matter, WM, and cerebrospinal fluid), and intensity modulation in SPM5 (Institute of Neurology, London, UK).

The Diffeomorphic Anatomical Registration Through Exponential
Lie Algebra toolbox was used for high-dimensional normalization.
For intensity modulation, voxel values from segmented images were multiplied by warped/unwarped measures derived from nonlinear spatial normalization, thereby converting relative regional gray matter density values into absolute gray matter density values (gray matter amount per unit brain tissue volume) prior to spatial normalization. Subsequently, 8-mm Gaussian kernel smoothing of the images was completed. Age, gender, education, psychiatric symptom severity, GAF scores, and MCCB scores were treated as covariates, and covariate influences were regressed out in multiple pattern recognition analysis. The threshold for significance was family-wise error rate corrected p < .05.

| DTI
All subjects underwent DTI concurrently with 3D-FSPGR. For DTI, a single-shot, spin-echo-planar sequence with a TR of 12,000 ms and a TE of 83.3 ms was conducted to generate 4-mm-thick slices (no gap) with a single excitation, 26-cm FOV, and spatial resolution of 1.02 mm × 1.02 mm × 4 mm. Diffusion properties were measured at b = 1,000 s/mm 2 along 25 noncollinear directions. Eddy current and motion corrections were performed on diffusion-weighted images with the Functional MRI of the Brain (FMRIB) Linear Image Registration Tool. Image distortion due to gradient nonlinearity image distortion was corrected for with "GradWarp." Individual functional anisotropy (FA) and mean diffusivity (MD) maps were calculated using the DTIFIT tool.
Because of inherent MRI geometric distortion FA map differences from T1/T2/proton density-weighted template images in SPM5, we created a FA template for the present study participants.
T2-weighted echo-planar image was coregistered in 3D-FSPGR images, and then, the coregistration parameter was applied to each corresponding FA map. The spatial normalization parameters from the 3D-FSPGR images were applied to coregistered FA maps. Normalized FA maps were smoothed with an 8 mm isotropic Gaussian kernel, thereby creating averaged images (FA templates). FA maps were transformed from native space to stereotactic space by registering each image in our FA template. Finally, smoothing of the normalized FA map was completed with an 8-mm isotropic Gaussian kernel.

| Image processing for TBSS
Voxel-wise statistical analysis was completed in TBSS version 1.2 software. We used the FMRIB Software Library (FSL, Oxford), including skull stripping and eddy current correction tools, to preprocess diffusion tensor images (FA, trace, axial and radial diffusivity).
Briefly, FA maps created for each subject using the FSL were aligned into a common (Montreal Neurologic Institute 152 standard) space with the nonlinear registration FSL tool FNIRT. All transformed FA images were averaged to create a mean FA image, and the tracts were narrowed to generate a mean FA skeleton incorporating the central white matter tracts common to all subjects. The voxel values of each subject's FA map were projected onto the skeleton. The FA threshold was set to 0.2 (TBSS default) to confine the analysis to white matter.
We conducted voxel-wise permutation-based nonparametric inference (Nichols & Holmes, 2002) on skeletonized the FA data using FSL Randomize version 2.1. Both patient > HC and patient < HC contrasts were identified with 5,000 permutations and a significance level of p < .05 (family-wise error rate corrected). We performed multiplecomparison corrections using threshold-free cluster enhancement (Smith & Nichols, 2009) to avoid arbitrary selection of the clusterforming threshold, while preserving the sensitivity benefits of cluster wise correction. To compare trace, axial diffusivity, and radial diffusivity, we used FA images with the FSL to achieve nonlinear registration and staged skeletonization, and to estimate projection vectors from each subject onto the mean FA skeleton.

| Statistical analyses
Sociodemographic and psychometric variables were compared among groups with LSD t tests. We carried out multivariate pattern   recognition analysis (Elton, Chanon, & Boettiger, 2019) to assess correlations between psychometric symptoms (i.e., AHRS, PANSS, HAMD, HAMA, and YMRS, MCCB, and GAF scores) and imaging data with a significance criterion of p < .05. Compared with the FUSCH-AVH group, the other five patient groups had specific characteristic differences ( Figure 3). Notably, the FUMDD-AVH group had greater GMVs in the posterior fron-

| Psychometric scores and correlation analysis
Mean AHRS, PANSS, HAMD, HAMA, CAPS, YMRS, and GAF scores and mean MCCB subscale scores obtained for each group are reported in Table 1 with the t and p values obtained for intergroup comparisons with LSD t tests. No GMV or FA/MD alterations were found to correlate significantly with any of the psychometrically assessed clinical symptoms.

| D ISCUSS I ON
To the best of our knowledge, this pilot study is the first study to report OTAVH-associated brain structural alterations across a number of mental disorder diagnosis groups of patients and HCs. This work is strengthened by our having enrolled only first-episode drugnaive patients to avoid pharmacological influences on the brain. The and occipital lobes also supports this disturbance hypothesis (Catani et al., 2011;Di Biase et al., 2019;EFL, 2016;McCarthy-Jones, Oestreich, & Whitford, 2015;Xie et al., 2019;Zhang et al., 2018).
Meanwhile, the enlarged striatal GMVs in our patients are consistent with the dopamine hypothesis of AVHs (Cassidy et al., 2018;Howes & Kapur, 2009;Russo et al., 2019). The present findings thus provide evidence potentially in support of two hypotheses of AVHs while providing clues for further research into OTAVH-specific features common to different mental disorders.
Second, we found that the FUBPD-OTAVH patient group had more marked alterations, relative to HCs, than the other patient groups. Contrasting our FUBPD-OTAVH and FUSCH-OTAVH comparisons with HCs after regressing out covariates (i.e., psychometric scores and sociodemographic variables), we found, surprisingly, that the FUBPD-OTAVH group had more widespread cortical GMV reductions and WM tract impairments. Given that there are quite limited data in the literature regarding GMV alterations and WM impairments in patients with borderline personality disorder and the available data show, for the most part, less pronounced alterations than those found in this study (Aguilar-Ortiz et al., 2018;Gan et al., 2016;Jin et al., 2016;Maier-Hein et al., 2014;Ninomiya et al., 2018;Rossi et al., 2015), it is difficult for us to speculate regarding an explanation for our findings in the FUBPD-OTAVH group. The present findings may thus provide important foundational information regarding divergent neurology in patients with borderline personality disorder who experience AVHs.
The differences appear to be inconsistent with a functional compensation mechanism because functional compensation would not be expected to be as consistent across individuals as differences observed in this study. Further research is needed to explain these alterations.
Altogether  independently. Sixth, we compared each diagnostic study group to every other diagnostic group in this study, a statistical practice that has been questioned for potential so-called "double dipping". In this regard, it is difficult to balance statistical conventions that prevent type I errors versus those that enable type II errors. In the current stage of our understanding, it is important to accumulate clues and thus this potential risk is taken with due caution. Seventh, although the patient populations addressed in this study often require higher dosage pharmacotherapies than similarly diagnosed patients without OTAVHs, these findings do not provide predictive prognostic information for patients. In the future, we will conduct a follow-up study to investigate how common, specific brain features in patients who experience OTAVHs are related to clinical implications and prognoses.

| CON CLUS ION
The present findings provide evidence consistent with the bottomup and top-down reciprocal action dysfunction hypothesis of AVHs F I G U R E 4 Cross comparisons of structural features among the FUBPD-, FUPTSD-, FUMN-, and FUMDD groups and the dopamine hypothesis of AVHs. We observed specific neurological features that provide clues for further research into OTAVHspecific features common to different mental disorders and potential treatment targets for AVHs.

ACK N OWLED G M ENTS
This work was supported by grants from the National Natural

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R S ' CO NTR I B UTI O N
CZhuo, CW, XS, XL, and YX conceived and designed research; CZhou, XX, YX, and CW collected data and conducted research; CZhuo, HT, DJ, WW, and GL analyzed and interpreted data; CZhou and CZhuo wrote the initial paper; CZhuo, WW, and CZhuo revised the paper; WW and HT had primary responsibility for final content. All authors read and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and analyzed during the present study are available from the corresponding author on reasonable request.