Neurological Soft Signs Predict Auditory Verbal Hallucinations in Patients With Schizophrenia

Abstract

Neurological soft signs (NSS) are well documented in individuals with schizophrenia (SZ), yet so far, the relationship between NSS and specific symptom expression is unclear. We studied 76 SZ patients using magnetic resonance imaging (MRI) to determine associations between NSS, positive symptoms, gray matter volume (GMV), and neural activity at rest. SZ patients were hypothesis-driven stratified according to the presence or absence of auditory verbal hallucinations (AVH; n = 34 without vs 42 with AVH) according to the Brief Psychiatric Rating Scale. Structural MRI data were analyzed using voxel-based morphometry, whereas intrinsic neural activity was investigated using regional homogeneity (ReHo) measures. Using ANCOVA, AVH patients showed significantly higher NSS in motor and integrative functions (IF) compared with non-hallucinating (nAVH) patients. Partial correlation revealed that NSS IF were positively associated with AVH symptom severity in AVH patients. Such associations were not confirmed for delusions. In region-of-interest ANCOVAs comprising the left middle and superior temporal gyri, right paracentral lobule, and right inferior parietal lobule (IPL) structure and function, significant differences between AVH and nAVH subgroups were not detected. In a binary logistic regression model, IF scores and right IPL ReHo were significant predictors of AVH. These data suggest significant interrelationships between sensorimotor integration abilities, brain structure and function, and AVH symptom expression.

Introduction

Dimension-based systems neuroscience might more accurately delineate disease pathophysiology and more readily suggest specific pharmacotherapeutic targets.^1–4 The number of studies that have examined fundamental dimensions of human functions in patients with mental disorders has dramatically increased in recent years.^5,6 As a consequence of this development, in 2019, the sensorimotor system dysfunction has been promoted as an essential addendum to the Research Domain Criteria (RDoC) framework.⁷ Since schizophrenia (SZ) patients with more severe sensorimotor symptom load (ie, parkinsonism and neurological soft signs [NSS]) show more pronounced clinical symptoms (ie, negative and positive symptoms) and poorer neurocognitive and psychosocial functioning,^8–11 this addendum to the RDoC matrix will help to decisively strengthen our understanding of how sensorimotor dysfunction mediates other dimensions, such as negative/positive valence systems, cognitive systems, or social processes in SZ.^12,13 Furthermore, overlapping circuitry is implicated in NSS and auditory verbal hallucinations (AVH), providing a plausible biological connection between sensorimotor and cognitive systems. For instance, NSS are linked to alterations of the inferior frontal gyrus, paracentral gyrus, inferior parietal lobe (IPL), bilateral putamen, cerebellum, and both the superior and middle temporal gyri (STG and MTG) (for an overview see also Hirjak et al^14,15 and Zhao et al¹⁶). The IPL, STG, and the transverse temporal gyrus (TTG; Heschl’s gyrus) also play a crucial role in the pathogenesis of AVH.^17,18 Sensorimotor deficits in the pathophysiology of AVH have been described in the form of altered self-monitoring and predictive coding, which lead to a misattribution of inner speech as disintegrated and externally generated.^19–21 According to the concept of “basic symptoms” (“Basissymptome”) presented by Huber and Gross,^22,23 neurochemical alterations might lead to aberrant information processing and to transphenomenal and unspecific basic disturbances. The so-called “basic symptoms” are a common intermediate link between neurobiological or neurophysiological alterations and flamboyant (manifest) symptoms of psychotic disorders.²⁴ This implies that a fundamental neurobiological process may lead to manifest symptoms of psychosis such as sensorimotor or auditory abnormalities. Until now, however, putative associations between NSS and positive symptoms, specifically AVH, have been hardly considered by previous research. In childhood-onset SZ, there is preliminary evidence that the interplay between sensorimotor and social-cognitive processing networks predicted the severity of positive symptoms, such as delusions and AVH.²⁵ Further preliminary support for a specific relationship between AVH and sensorimotor dysfunction exists from a recent study demonstrating that individuals with high schizotypy and high AVH predisposition show significantly higher NSS compared with individuals without schizotypal traits.²⁶

Here, we selectively tag this relationship in SZ by integrating well-established clinical/behavioral and biological measures and analyzing both constructs dimensionally, along a spectrum of functioning. We used psychometric data in conjunction with a detailed examination of sensorimotor function and multimodal magnetic resonance imaging (MRI) to investigate associations between NSS, brain structure and function, and AVH. In particular, we chose to investigate regional homogeneity (ReHo), since this method provides an excellent trade-off between signal and noise in the data, as much as it has been shown to be a highly sensitive, reproducible, and reliable neuroimaging marker to characterize the human brain function according to its spatial organization, as well as its relationship with other network-level features.^27–29 Unlike the fractional amplitude of low-frequency fluctuations approach (fALFF; signal variability of a single voxel in the low frequency domain), ReHo measures the synchronicity of the time series between a given voxel and each of its neighboring voxels.^27,28 This approach has been increasingly used in previous rs-fMRI studies in SZ, including research on NSS and AVH.^27,30–33

We expected that significant relationships between NSS and psychotic symptoms would particularly apply to the sensory domain, ie, such associations would essentially pertain to AVH and not to delusions. More specifically, we hypothesized that AVH severity is associated with the NSS domain integrative functions (IF), since this sensorimotor domain specifically reflects the integration of both sensory and motor functions.³⁴ In addition, the IF domain can very well map complex interactions between sensorimotor and higher-order integrative cognitive functions. Finally, using a region-of-interest (ROI) approach and logistic regression models, we hypothesized that NSS scores, gray matter volume (GMV), and intrinsic neural activity of specific brain regions that have been previously associated with NSS IF domain^34,35 and AVH,^18,36 respectively, such as IPL, STG, MTG, and paracentral lobule would predict AVH severity in SZ.

Methods

Participants

We examined a total of 76 consecutive cases treated at the Department of Psychiatry and Psychotherapy at the Central Institute of Mental Health in Mannheim, Germany (see table 1 for detailed demographics and clinical scores) and meeting DSM-IV-TR³⁷ criteria for SZ. Inclusion and exclusion criteria are listed in the supplement. The local Ethics Committee (Medical Faculty at Heidelberg University, Germany) approved the study. Written informed consent was obtained from all SZ patients after all aims and procedures of the study had been fully explained.

Table 1.

Demographics and Clinical Scores for Schizophrenia Patients With Auditory Verbal Hallucinations (AVH) and Without Auditory Verbal Hallucinations (nAVH)

Variable	nAVH (n = 34) mean ± SD	AVH (n = 42) mean ± SD	t  a	df	P
Age (y)	37.09 ± 9.52	40.29 ± 13.12	−1.19	74	.24
Gender (m/f)b	18/16	22/20	0.002	1	.96
Education (y)	13.03 ± 2.17	12.43 ± 2.37	1.14	74	.26
Duration of illness (y)	9.18 ± 9.34	13.79 ± 11.44	−1.89	74	.062
OLZ	14.48 ± 9.28	21.33 ± 10.48	−2.98	74	.004
PANSS
Positive	13.06 ± 6.15	19.02 ± 7.19	−3.83	74	<.001
Negative	15.15 ± 8.31	19.05 ± 7.11	−2.20	74	.031
General	30.06 ± 8.03	40.93 ± 9.74	−5.22	74	<.001
Total	58.26 ± 17.50	79.00 ± 16.90	−5.23	74	<.001
BPRS
AVH	1± 0	3.93 ± 1.52	−11.22	74	<.001
Delusions	5.53 ± 2.95	8.64 ± 3.75	−3.95	74	<.001
Total	31.29 ± 10.00	45.05 ± 10.20	−5.89	74	<.001
NSS
IF	1.94 ± 1.10	3.19 ± 1.76	−3.61	74	.001
MOCO	6.65 ± 3.50	9.14 ± 4.12	−2.80	74	.006
COMT	3.12 ± 2.47	3.71 ± 2.19	−1.12	74	.27
R/L/S/O	2.35 ± 2.41	3.26 ± 3.05	−1.42	74	.16
HS	2.82 ± 1.64	3.23 ± 1.96	−0.98	74	.33
NCRS total	2.56 ± 3.40	3.43 ± 2.99	−1.18	74	.24
BARS global	0.44 ± 0.86	1.36 ± 1.53	−3.118	74	.003
AIMS total	0.50 ± 1.80	1.59 ± 2.64	−2.06	74	.043

Note: SD, standard deviation; df, degrees of freedom; OLZ, olanzapine equivalents; PANSS, Positive and Negative Syndrome Scale; BPRS, Brief Psychiatric Rating Scale; NSS, neurological soft signs, as examined with the Heidelberg Scale; IF, integrative function; MOCO, motor coordination; COMT, complex motor task; R/L/S/O, right/left and spatial orientation; HS, hard signs; NCRS, Northoff Catatonia Rating Scale; BARS, Barnes Akathisia Rating Scale; AIMS, Abnormal Involuntary Movement Scale.

^aThe t-values were obtained using a 2-tailed independent samples t-test.

^bThe P-values for distribution of gender were obtained by a chi-square test.

Clinical Assessment

Patients were recruited and examined within 1 week after reaching partial remission defined as >25% reduction of symptoms according to the Positive and Negative Syndrome Scale (PANSS) total score³⁸ to ensure optimal assessment conditions. The duration between enrollment and participation consent, evaluation of psychopathology (PANSS,³⁸ Brief Psychiatric Rating Scale [BPRS]),³⁹ sensorimotor assessment (Heidelberg Neurological soft signs Scale [NSS]⁴⁰; see supplementary material for NSS subscales, Simpson and Angus Scale [SAS],⁴¹ Northoff Catatonia Rating Scale [NCRS],⁴² and Abnormal Involuntary Movement Scale [AIMS]),⁴³ and MRI examination was less than 3 days. We have chosen these particular scales to make the results of the present study comparable with other studies on SZ patients, including own previous work as much as previous independent research. Participants were dichotomized based on whether or not they were experiencing AVH. AVH severity was determined using the corresponding BPRS item 12. For measuring delusions (regardless of specific delusional content), the sum of BPRS items 8, 11, and 15 was considered.

MRI Data Acquisition

MRI acquisition took place at the Central Institute of Mental Health, Mannheim, Germany, on a 3-Tesla MAGNETOM TIM Trio MR scanner (Siemens Medical Systems, Erlangen, Germany) equipped with a 32-channel multi-array head coil. The scanner protocol included 3 measurements: a resting-state scan and 2 structural scans (diffusion-tensor-imaging data and 3-dimensional magnetization-prepared rapid gradient-echo [3D-MPRAGE] images). Technical details on structural and functional MRI sequences are provided as supplementary information.

MRI Data Processing

Voxel-based morphometry (VBM) of T1-weighted structural images was calculated for gray matter density in each voxel using CAT12 (http://dbm.neuro.uni-jena.de/cat/, version 12.6 (r1450, April 4, 2019), which is an extension for SPM12 (http://www.fil.ion.ucl.ac.uk/spm/). CAT12 default settings were used for VBM. The processing pipeline included (1) segmentation of images into gray matter, white matter, and cerebrospinal fluid using a voxel size of 1.5 mm³; (2) normalization using the DARTEL approach⁴⁴; and (3) smoothing the gray matter probability values using an 8-mm full-width half-maximum (FWHM) isotropic Gaussian kernel. ReHo analysis^28,45,46 was applied to rs-fMRI images using the Data Processing Assistant for rs-fMRI (DPARSF, version 3.1)⁴⁷ running in MATLAB R2016a. To facilitate comparisons with the extant literature, we chose a cluster size of 27 voxels and ReHo smoothing kernel of 4 mm, as recommended by the DPARSF authors and as employed in several rs-FMRI studies using ReHo measures in attenuated-psychosis syndrome⁴⁸ and SZ.^49–51 Details on ReHo pipeline are provided as supplementary information.

Statistical Analysis

We first tested whether nAVH and AVH groups significantly differed with respect to age, gender, education, olanzapine equivalents (OLZ), BPRS (AVH, delusions, and total scores), and NSS (IF, motor coordination [MOCO], complex motor task [COMT], right/left and spatial orientation [R/L/S/O], and hard signs [HS]). To test for potential between-group differences considering catatonia symptoms, the NCRS total score was used. Next, using partial correlations adjusted for age, gender, education, OLZ, NCRS total, BARS global score, and AIMS total scores, we investigated associations between NSS and AVH severity, as well as between NSS and severity of delusions. For completeness, using ANCOVAs adjusted for age, gender, education, OLZ, NCRS total, BARS global, and AIMS total scores, we investigated between-group differences in brain structure and function considering the 4 aforementioned ROIs. Data normality was investigated using the Kolmogorov–Smirnov test and homogeneity using Levene’s test, respectively.

Finally, we used a binary logistic regression model⁵² to examine whether NSS scores, GMV, and ReHo in the 4 ROIs could predict AVH. Age, OLZ, NCRS total, BARS global, and AIMS total scores were included as covariates. The statistical significance of individual regression coefficients was evaluated using the Wald chi-square statistic. Hosmer-Lemeshow test was used to evaluate the model’s goodness-of-fit.^52,53 The pseudo R square obtained by Cox and Snell⁵⁴ and Nagelkerke⁵⁵ methods was used as an additional descriptive measure of goodness-of-fit. Validations of predicated probabilities were evaluated by a classification table. The cutoff point was set at 0.5. A classification table was used to measure sensitivity (the ability to correctly predict patients with AVH) and specificity (the ability to correctly predict patients with nAVH).

Additionally, we used a hierarchical multiple regression model to test if NSS scores, GMV, and ReHo (adjusted for age, OLZ, NCRS total, BARS global, and AIMS total scores) predicted delusions severity. Prior to conducting the regression, the assumptions of this statistical analysis were tested. As the tolerance statistics for collinearity were all above the accepted threshold (all tolerance values > 0.66), the assumption of multilinearity was deemed to have been met. Residual and histogram plots indicated that the assumptions of normality and linearity were satisfied. For the standardization of independent variables of the regression models, each variable was divided by 2 times its standard deviation.⁵⁶ For all analyses, a nominal significance threshold of P < .05 was defined. Where applicable, a correction for multiple comparisons using the false discovery rate (FDR⁵⁷) was used. Statistical analyses were performed using MATLAB version R2016a and SPSS version 24.

Results

There was a significant difference between mean NSS IF scores for the nAVH (mean ± SD: 1.94 ± 1.10) and AVH (3.19 ± 1.76) groups, t(74) = −3.61, P = .001, d = 0.85. Also, a significant difference between MOCO scores of nAVH (6.65 ± 3.50) and AVH (9.14 ± 4.12) groups was observed, t(74) = −2.80, P = .006, d = 0.65. Other NSS subscores (COMT, R/L/S/O, and HS) were not significantly different across groups, all P-values >= .16 (table 1). At the time of inclusion, most of the patients were receiving treatment with antipsychotic agents according to their psychiatrists’ choice. A partial correlation (2-tailed) was run to determine the relationship between an individual’s OLZ and AIMS total score, BARS global score, and NSS IF scores while controlling for age and gender. There was no significant partial correlation between OLZ and sensorimotor symptoms in the nAVH (all P-values > .29) or AVH (all P-values > .062) patients.

In partial correlation analyses, NSS IF scores were significantly positively associated with AVH severity, r = .360, P = .02, FDR-corrected. This was not the case in analyses that tested for associations between NSS and delusion severity, r = .027, P = .82, FDR-corrected. No other significant correlations were observed. More details are provided in table 2. In ROI-based analyses, GMV and ReHo in the paracentral lobule, IPL, MTG, and STG did not significantly differ between AVH and nAVH patients, all P-values >= .07, FDR-corrected. For further details, see table 3.

Table 2.

Partial Correlation Between AVH, Delusions, and NSS Subscores Adjusted for Age, Gender, Education, OLZ, NCRS Total, BARS Global, and AIMS Total Scores (n = 76)

	IF	MOCO	COMT	R/L/S/O	HS
AVH
Correlation	0.360	0.194	0.073	0.105	−0.078
Significance	0.02	0.37	0.68	0.68	0.68
Delusions
Correlation	0.027	0.054	−0.192	0.095	−0.095
Significance	0.82	0.72	0.37	0.68	0.68

Note: NCRS, Northoff Catatonia Rating Scale; BARS, Barnes Akathisia Rating Scale; AIMS, Abnormal Involuntary Movement Scale; AVH, auditory-verbal hallucinations; IF, integrative function; MOCO, motor coordination; COMT, complex motor task; R/L/S/O, right/left and spatial orientation; HS, hard signs; FDR, false discovery rate. P-values are FDR-corrected. Two-tailed significant P-values (P < .05) are marked in bold.

Table 3.

Comparisons Between GMV and ReHo Means in 4 Regions of Interests Across Schizophrenia Groups With (AVH) and Without (nAVH) Auditory Verbal Hallucinations

	nAVH (n = 34) mean ± SD	AVH (n = 42) mean ± SD	F  a	P  b	95% C.I.		Partial η2
					Lower	Upper
GMV
Left middle temporal gyrus	12.97 ± 1.74	12.25 ± 1.89	0.11	.74	−0.67	0.93	0.002
Left superior temporal gyrus	5.8 ± 0.78	5.86 ± 0.89	0.29	.59	−0.49	0.28	0.004
Right inferior parietal lobule	8.33 ± 0.89	8.09 ± 1.22	0.021	.89	−0.51	0.45	<0.001
Right paracentral lobule	1.67 ± 0.22	1.65 ± 0.23	0.0003	.99	−0.12	0.12	<0.001
ReHo
Left middle temporal gyrus	0.94 ± 0.03	0.93 ± 0.02	0.09	.77	−0.02	0.01	0.001
Left superior temporal gyrus	0.93 ± 0.04	0.92 ± 0.04	0.01	.91	−0.02	0.02	<0.001
Right inferior parietal lobule	1.09 ± 0.02	1.07 ± 0.02	3.27	.07	−0.001	0.03	0.047
Right paracentral lobule	1.04 ± 0.05	1.05 ± 0.05	0.31	.58	−0.03	0.02	0.005

Note: NCRS, Northoff Catatonia Rating Scale; BARS, Barnes Akathisia Rating Scale; AIMS, Abnormal Involuntary Movement Scale; C.I., confidence interval; GMV, gray matter volume (in mm³); ReHo, regional homogeneity; FDR, false discovery rate.

^aAnalysis of variance (ANOVA). Age, gender, education, OLZ, NCRS total, BARS global, and AIMS total scores were used as covariates in the model.

^b  P-values are FDR-corrected for multiple comparisons.

Overall, model evaluation showed that the logistic model was a better fit to the data than the null model, X²(15, n = 76) = 34.10, P = .002. The model revealed that NSS IF scores and right IPL ReHo were significant predictors of AVH. See table 4 for statistical details and figure 1 for the anatomical structure of the IPL. The model was fit to the data well validated by the Hosmer-Lemeshow test, X²(8) = 7.18, P = .52. The pseudo R square obtained by Cox and Snell and Nagelkerke methods was 0.38 and 0.51, respectively. The classification table revealed that the model had a sensitivity of 76.2% and a specificity of 82.3% (table 5). NSS IF and right IPL ReHo were significant predictors of AVH, but GMV was not. Further details are provided in supplementary table 1. To further expand on potential contributions of other brain regions to both NSS and AVH, we extended the initial ROI selection to the primary motor and auditory cortex (the TTG). Bilateral mean ROI values were extracted from precentral gyrus and TTG using Neuromorphometrics atlas definitions. These regions were not significant predictors of AVH (see supplementary table 1).

Table 4.

Binary Logistic Regression Analysis of 76 Schizophrenia Patients With and Without AVH

Variables	B	S.E.	Wald’s X2	df	Sig.	odds ratio (95% C.I.)
Age	1.84	0.94	3.84	1	0.05	6.27 (1.00 to 39.41)
OLZ	1.78	0.75	5.67	1	0.017	5.96 (1.37 to 25.90)
NCRS total score	−0.14	0.69	0.02	1	0.84	0.87 (0.23 to 3.34)
BARS global score	2.52	1.17	4.62	1	0.032	12.40 (1.25 to 123.25)
AIMS total score	−0.66	0.92	0.52	1	0.47	0.51 (0.09 to 3.10)
NSS
IF	2.01	0.90	4.92	1	0.027	7.42 (1.26 to 43.63)
MOCO	0.76	0.85	0.81	1	0.37	2.15 (0.40 to 11.41)
GMV
Left middle temporal gyrus	−0.07	1.47	0.002	1	0.96	0.93 (0.05 to 16.65)
Left superior temporal gyrus	0.38	1.05	0.13	1	0.72	1.46 (0.18 to 11.51)
Right inferior parietal lobule	0.84	1.45	0.33	1	0.56	2.31 (0.13 to 39.62)
Right paracentral lobule	0.60	0.80	0.57	1	0.45	1.83 (0.38 to 8.79)
ReHo
Left middle temporal gyrus	0.79	0.86	0.83	1	0.36	2.19 (0.40 to 11.88)
Left superior temporal gyrus	0.47	0.88	0.29	1	0.59	1.61 (0.29 to 8.10)
Right inferior parietal lobule	−1.88	0.92	4.18	1	0.041	0.15 (0.02 to 0.92)
Right paracentral lobule	−0.04	0.68	0.004	1	0.95	0.96 (0.25 to 3.67)

Note: AVH, auditory verbal hallucinations; OLZ, olanzapine equivalents; NCRS, Northoff Catatonia Rating Scale; BARS, Barnes Akathisia Rating Scale; AIMS, Abnormal Involuntary Movement Scale; NSS, Neurological soft signs; IF, integrative functions; MOCO, motor coordination; GMV, gray matter volume; ReHo, regional homogeneity; S.E., standard error of the mean; C.I., confidence interval. Significant P-values (P < .05) are marked in bold.

Fig. 1.

Anatomical location of 4 regions of interests (ROIs) highlighted in red, including left STG and MTG, right IPL, and right paracentral lobule. Note: STG, superior temporal gyri; MTG, middle temporal gyri; IPL, inferior parietal lobule.

Table 5.

Classification of the Observed and Predicted Frequencies for AVH by Logistic Regression (see also Table 4)

Observed	Predicted
	nAVH	AVH	% Correct
nAVH	28	6	82.3
AVH	10	32	76.2
Overall % correct			78.9

Note: nAVH, schizophrenia patients without auditory verbal hallucinations; AVH, schizophrenia patients with auditory verbal hallucinations. The cutoff was set to 0.50. Sensitivity = 32/(10 + 32) = 76.2%; specificity = 28/(28 + 6) = 82.3%; false positive = 6/(6 + 32) = 15.8%; false negative = 10/(28 + 10) = 26.3%.

A hierarchical multiple regression was used to test whether delusion severity can also be predicted by NSS scores, GMV, and ReHo activities. Age, OLZ, NCRS total, BARS global, and AIMS total scores were entered as covariates in the first step, and the rest of the variables in the second step. The regression model revealed that at the step one, age, OLZ, NCRS, BARS, and AIMS scores did not contribute significantly to the regression model, F (5, 70) = 0.96, P = .45). Introducing NSS IF scores, GMV, and ReHo indexes of 4 brain regions did not significantly change the explained variance, F(15, 60) = 0.67, P = .80. The partial correlation (2-tailed; n = 76) between NSS performance and delusions adjusted for age, gender, education, OLZ, NCRS total, BARS global, and AIMS total scores did not reveal any significant association (supplementary table 2).

Discussion

We used psychometric data in conjunction with a detailed examination of sensorimotor function and multimodal MRI to investigate relationships between NSS, AVH, and GMV and intrinsic neural activity in SZ patients. Three main findings emerged: First, specific NSS subdomains (ie, IF and MOCO) significantly differed between AVH and nAVH patients, and IF scores were significantly associated with AVH severity (see figure 2). Second, such relationships were not detected for delusions. Third, in logistic regression models, IF scores and ReHo of the right IPL were significant predictors of AVH. Such relationships were not detected for delusions.

Fig. 2.

Scatter plots of partial correlations between NSS IF and BPRS delusions (composite) scores, r = .365, P = .01 (left), and between NSS IF and BPRS AVH scores, r = .017, P = .89 (right). P-values are FDR-corrected. Note: NSS, neurological soft signs; IF, integrative functions; BPRS, Brief Psychiatric Rating Scale; AVH, auditory verbal hallucinations; FDR, false discovery rate.

Association Between NSS and AVH

The first finding that SZ patients with higher IF scores show more severe AVH than SZ patients with a lower level of sensorimotor abnormalities extends our current understanding of interrelated behavior in the RDoC Matrix. Numerous studies have demonstrated that individuals with higher NSS exhibit more severe positive symptoms.^40,58,59 This said, SZ patients showing difficulties during the correct execution of station and gait and tandem walking, which is reflected by higher NSS IF scores, exhibit a failure to monitor movement (proprioceptive) experiences that are self-generated.⁶⁰ In terms of AVH, aberrant monitoring of inner (self) speech can also lead to alterations of agency.^60,61 It is evident that disturbance of agency might be involved in 2 different clinical phenotypes such as NSS and AVH. This is also supported by the second finding, as we did not found a significant association between NSS scores and other psychotic symptoms such as delusions (see table 2). This indicates that relationships between NSS and AVH are essentially driven by sensorimotor/auditory abnormalities, in contrast to aberrant (lateral prefrontal) belief evaluation systems that have been suggested to underlie delusion formation.^62–64 Furthermore, in the present study, we found a strong correlation between the right IPL GMV and the left STG GMV in all patients, r = .75, P < .001, n = 76. Interestingly, the correlation between these 2 regions was significantly stronger in AVH patients, r = .89, P < .001, n = 34, than in nAVH patients, r = .63, P < .001, n = 42, tested by Fisher’s z-transformation, z = −1.72, P = .043. This finding is not surprising given the wealth of neuroimaging evidence on STG disturbance in SZ.^65–67 Numerous MRI studies^{21,60,68–70} and meta-analyses^67,71–74 have demonstrated that disrupted neural circuits consisted of STG and involved in the generation, monitoring, and perception of speech are most likely to explain the development of AVH in SZ. The strong correlation between the right IPL GMV and left STG GMV fits well with a former structural study,⁷⁰ as well as to our previous data on structural networks in 2 independent samples.^36,75 This finding could also reflect the importance of deficits in self-monitoring of inner speech as indicated by cognitive models on AVH^74,76 and experiments with an involvement of the (right) temporoparietal junction.⁷⁷ Accordingly, another study in SZ patients with AVH showed an aberrant functional connectivity between a speech-related system and a network subserving self-referential processing.⁷⁸

The IPL (Brodmann’s area 40) is responsible for the immediate guidance of our bodily actions in space,^79,80 organization, and control of visuomotor acts.⁸⁰ Interestingly, structural alterations of the IPL were also associated with higher NSS scores in SZ patients, organization, and control of visuomotor acts.^34,80,81 On the other hand, STG is mainly involved in managing both object- and space-related information. The significant interplay between the right IPL (region of the sensorimotor network) and the left STG (region of the auditory network) is also in line with a number of neuroimaging studies that showed an interaction (functional coupling) between sensorimotor and auditory regions while listening to speech,^82–84 tools sounds,⁸⁵ music,⁸⁶ and rhythms.⁸⁷ A more recent magnetoencephalography study⁸⁸ found that motor and auditory cortices are synchronized for a restricted range of frequencies (~4.5 Hz). These studies and the results of the present study suggest a close relationship of motor and auditory cortex, ie, structural and functional interactions between sensorimotor domain (NSS subscale IF) and psychopathological symptoms (AVH). Dysfunctional self-monitoring, eg, the misattribution of self-generated actions as externally generated, has been suggested as a fundamental mechanism in SZ^89,90 and particularly in AVH.^60,61

Specifically, different models suggest a disturbance between top-down and bottom-up processes^60,76 with respect to deficits in self-monitoring of inner speech^21,91 and predictive coding,²⁰ eg, a disconnection between sensory and motor processing leading to a mismatch between the predicted auditory consequences of self-generated inner speech and the current auditory signal.^92,93 Decreased activity of the somatosensory cortex was associated with both impaired sensory prediction and AVH severity.^36,95 Further evidence exists on a disruption between neural networks representing the so-called “intrinsic” and “extrinsic” self-aspects, ie, self-referential processing and the dynamic interaction with the external world.⁹⁴ In AVH, “resting-state hypothesis” postulates an abnormal resting-state activity in one of the sensorimotor core regions of AVH, ie, the temporal cortex and deficits in its rest-rest communication with the default mode network (DMN) and rest-stimulus interaction with external stimuli.⁹⁵At this point, the consideration of different symptom dimensions (such as NSS, catatonia, and AVH) of SZ is of particular interest, regarding a common feature in a miscommunication between, particularly, the DMN, the salience network, and sensorimotor networks.^78,96,97 In the present study, however, we followed a more integrative approach to study the interplay between NSS IF, different sensorimotor core regions, and AVH.

Prediction of AVH

The third finding highlights the crucial role of the IPL in AVH. Furthermore, it proposes a model composed of sensorimotor assessment and MRI data that is able to predict AVH with a sensitivity of 76.2% and a specificity of 82.3%. This finding is interesting for several reasons: First, our data support a relationship between NSS IF scores, related key regions, and the existence of AVH. One explanation could be that higher NSS levels may particularly apply to a subgroup of SZ patients with more severe disease course in general, as NSS has been associated with a poor response to antipsychotics in SZ patients and treatment resistance.⁹⁸ In approximately 25% of SZ patients, AVH are refractory to psychotropic drug treatment and can chronically persist.⁹⁹ Second, these data suggest that observable and measurable behavior, such as IF score and AVH, share neuromechanisms in terms of disrupted structural and functional brain circuits. It is important to note that the IPL plays a crucial role in both cognitive and sensorimotor deficits leading to different clinical phenotypes in terms of AVH and NSS.¹⁰⁰ Third, NSS are observable before first-episode psychosis, as much as during the disease course, and their neurodevelopmental origin is obvious.^35,101 Finally, we show for the first time that clinical assessment of sensorimotor abnormalities together with morphological alterations of the NSS-associated regions (ie, the IPL) are able to predict the clinical severity of AVH in SZ. This is potentially relevant for clinical routine, since reliable objective predictors of specific psychopathology are still missing.

Strengths and Limitations

The strengths of this study include the sample size, the hypothesis-driven approach, and the comprehensive psychometric and clinical assessment in conjunction with the multimodal MRI approach. Nevertheless, this study had clear limitations, such as the cross-sectional design and patients’ antipsychotic medication. Another limitation of this study is the overlap of different sensorimotor symptoms and signs. Although we have used established clinical rating scales, we further acknowledge a dimensional view of sensorimotor dysfunction in psychiatric disorders rather than categorical classification. Furthermore, the AVH group may reflect clinical and psychometric heterogeneity, as we did not discriminate between different types of episodic and persistent symptoms and treatment-resistant AVH. In future studies, it would be essential to investigate if the same regions are also found on a transdiagnostic continuum, eg, in healthy populations with AVH proneness as well as in other mental disorders with manifest AVH and NSS. This procedure could identify transdiagnostic mechanisms of AVH/NSS and positively influence the current discussion of categorical classification of mental disorders.

Conclusions

Given that alterations of the right IPL ReHo as well as NSS scores are able to predict AVH in SZ patients, this study provides the first evidence for a common pathophysiological mechanism that could lead to different clinical phenotypes such as NSS and AVH. Our findings support the idea of aberrant observable behavior and neuroimaging findings in the sensorimotor domain leading to AVH in SZ. A deeper understanding of the extent to which sensorimotor dysfunction measured by NSS and associated (sensori)-motor networks predict AVH could contribute to a better approach of individualized treatment options and to more accurate prediction in individuals at ultra-high risk for psychosis.¹⁴

Funding

This work was supported by the German Research Foundation (DFG, grant number DFG HI 1928/2-1 to D.H.; WO 1883/2-1 and WO 1883/6-1 to R.C.W.).