Cerebral hemodynamic characteristics of patients with auditory verbal hallucinations and the construction of nomogram models

Abstract

BACKGROUND

The characteristics of cerebral hemodynamic indexes of patients with different types of auditory verbal hallucinations (AVHs) was not clear.

AIM

To explore the characteristics of cerebral hemodynamic indexes of patients with different types of AVHs and construct the risk nomogram prediction model of patients with different types of AVHs.

METHODS

Patients with different types of verbal hallucinations who visited Wenzhou Seventh People’s Hospital were retrospectively selected from March 2021 to March 2023, and these patients were classified into 117 cases of schizophrenia (SCZ) with AVHs, 108 cases of post-traumatic stress disorder (PTSD) with AVHs, and 105 cases of recurrent depressive disorder with AVHs according to type. Transcranial doppler was performed to measure the hemodynamic parameters of the anterior cerebral artery (ACA), middle cerebral artery (MCA), posterior cerebral artery (PCA), basilar artery (BA) and vertebral artery (VA). Logistic regression modelling was used to explore the factors affecting patients with different types of AVHs and odds ratio, 95% confidence interval (CI). A clinical prediction model was constructed, and the efficacy of the clinical prediction model was evaluated by using receiver operating characteristic, Hosmer-Lemeshow Goodness-of-Fit test, calibration curves and decision curve analysis.

RESULTS

The differences between the three groups of patients in mean velocity (Vm)-MCA, end-diastolic velocity (Vd)-MCA, Vm-ACA, pulsatility index (PI)-ACA, Vm-PCA, peak systolic velocity (Vs)-PCA, Vd-PCA, Vm-BA, Vs-BA, Vd-BA, PI-BA, resistance index (RI)-BA, Vm-VA, Vs-VA, Vd-VA, PI-VA, and RI-VA indexes were statistically significant. Rising Vm-ACA is an independent risk factor for SCZ with AVHs, and falling Vm-VA, Vd-MCA, and Vd-VA are independent risk factors for SCZ with AVHs. Rising Vm-ACA, Vm-PCA, Vs-PCA, Vd-PCA, Vm-BA, and Vs-BA are independent risk factors for PTSD with AVHs, and Vm-MCA, Vs-MCA, Vd-MCA, PI-PCA, and RI-BA are independent protective factors for PTSD with AVHs. Elevated Vm-MCA, Vd-MCA, RI-BA, Vm-VA, and Vd-VA were independent risk factors, and elevated Vm-ACA, Vs-ACA, Vm-PCA, Vs-PCA, and Vd-PCA were independent protective factors. The areas under the curve of the three models were 0.82 (95%CI: 0.76-0.87), 0.88 (95%CI: 0.83-0.92), and 0.81 (95%CI: 0.77-0.86), respectively; the Hosmer-Lemeshow Goodness-of-Fit test of the calibration curves of the three models suggests that P > 0.05.

CONCLUSION

Monitoring the cerebral hemodynamic indexes of patients with AVHs is of practical significance in determining the type of mental disorder, which helps clinicians identify the type of AVHs and adopt more efficient treatment strategies to help patients recover.

Key Words: Auditory verbal hallucinations; Schizophrenia; Posttraumatic stress disorder; Recurrent depressive disorder; Nomogram model

Core Tip: This study explores cerebral hemodynamic characteristics in patients with auditory verbal hallucinations (AVHs) across schizophrenia, post-traumatic stress disorder, and recurrent depressive disorder. Key findings include increased mean velocity-anterior cerebral artery and end-diastolic velocity-vertebral artery as independent risk factors for AVHs in schizophrenia, while decreased end-diastolic velocity-resistance index (middle cerebral artery) and mean velocity-vertebral artery serve as protective factors. Nomogram models were developed to predict the risks of AVHs, demonstrating high discriminative ability with an area under the curve greater than 0.80 and significant clinical utility, thereby supporting early diagnosis and intervention strategies.

INTRODUCTION

Auditory verbal hallucinations (AVHs) are experiences that produce verbal auditory perceptions in the absence of corresponding external stimuli. They are part of a continuum that exists from healthy populations to patients with mental disorders. Epidemiological studies have demonstrated that the prevalence of AVHs is approximately 70% in individuals diagnosed with schizophrenia (SCZ), 10%-23% in those with recurrent depressive disorder (RDD), 23% in those with bipolar disorder, 46% in those with borderline personality disorder, 50% in those with post-traumatic stress disorder (PTSD), and 10%-20% in the general population[1,2]. Symptoms of AVHs are associated with numerous adverse outcome indicators and can cause severe distress, depression, impaired or even loss of social functioning, and increased risk of suicide in hallucinating individuals, which can have a significant impact on the patients themselves and their families.

In accordance with the tenets of neuropsychological theory, AVHs may be attributable to a discrepancy between the internal speech that is perceived and the attributions that are ascribed to it. Internal speech is defined as speech activity that is not voiced in response to thought activity. This discrepancy may be attributed to deficiencies in the patient’s internal speech processing or a lack of perception or understanding of internal speech to some degree[3]. Current research suggests that the symptoms result from a combination of abnormalities in the temporal lobe, the thalamus, and endocrine gland function. It is linked to activation of the posterior superior temporal gyrus and the transverse temporal gyrus, with severity inversely proportional to superior temporal gyrus volume[4,5]. The treatment of AVHs encompasses three principal domains: Pharmacological intervention, physiotherapy, and psychotherapy. Among pharmacological treatments, antipsychotics are the preferred option, as they provide relatively rapid symptom control. However, it should be noted that not all patients benefit from them, and many suffer from poor medication adherence[6]. The mainstay of physiotherapy is the use of transcranial direct current stimulation and repetitive transcranial magnetic stimulation[7,8]. However, the therapeutic efficacy of these techniques is highly variable, and the optimal parameters for stimulation, including the stimulation site, intensity, and duration, have yet to be established. Cognitive behavior therapy for psychosis is the main method of psychotherapy recommended by the relevant guidelines, but the scope of application is often limited by the patient’s economic conditions, the degree of knowledge of the disease, and potential stigmas associated with psychotherapy[9].

Prior research has demonstrated that patients with AVHs exhibit distinctive alterations in blood flow, with disruptions in blood flow distribution correlating with the clinical manifestations of AVH symptoms. Changes in functional brain activity can change hemodynamics. In the presence of AVHs, there may be changes in blood flow rate to the brain in the relevant area, such as an increase or decrease[10]. Cerebral hemodynamics is a diagnostic method that measures blood flow velocity in intracranial arteries. This is primarily achieved through Doppler ultrasound flow imaging, which can detect the presence of stenosis and occlusion of cerebral blood vessels in patients. Changes in cerebral hemodynamic indexes are closely associated with functional brain activities. This research sought to retrospectively examine alterations in cerebral blood flow patterns among patients experiencing various types of AVHs[11]. Additionally, it aimed to develop a nomogram model through logistic regression analysis to offer evidence-based support for the clinical assessment of AVH types.

MATERIALS AND METHODS

Patient population

The research subjects were selected from a cohort of AVH patients admitted to the Wenzhou Seventh People’s Hospital between March 2021 and March 2023. The subjects were divided into three groups based on their disease type: SCZ with AVHs (SCZ group), PTSD with AVHs (PTSD group), and RDD with AVHs (RDD group). The inclusion criteria for patients were: (1) Meeting the diagnostic criteria for SCZ, PTSD or RDD in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition; (2) The diagnosis of AVHs was consistent with the evaluation criteria of the Chinese version of the Psychiatric Symptom Rating Scale; (3) Age 18-60 years old; and (4) Informed consent to the study. Exclusion criteria were as follows: (1) Patients with previous epilepsy, head trauma and other disorders of consciousness; (2) Patients with severe organic brain diseases, chronic diseases, and systemic diseases; (3) Patients with other mental disorders; (4) Women during pregnancy or breastfeeding; (5) Patients with serious drug side effects; and (6) Patients with a history of electroconvulsive therapy or contraindications to electroconvulsive therapy or magnetic resonance imaging scanning within 6 months prior to enrollment. A total of 330 patients were included in this study following screening according to the pre-established inclusion and exclusion criteria. The sample was composed of 117 patients in the SCZ group, 108 patients in the PTSD group, and 105 patients in the RDD group. The recruitment procedure is detailed in Figure 1. This study was conducted in strict accordance with the ethical guidelines of the Declaration of Helsinki and approved by the Ethics Committee of the Wenzhou Seventh People’s Hospital. All patients voluntarily signed informed consent forms.

Figure 1 Flow chart of the study. PTSD: Post-traumatic stress disorder; RDD: Recurrent depressive disorder.

Collection of clinical data

The following general data were collected: gender, age, years of education and course of disease.

Cerebral hemodynamic indexes

Enrolled patients received transcranial doppler tests: (1) The patient was seated and lying with their back neck exposed. The operator stood to the left of the patient and used a 2-M doppler probe to obtain blood flow parameters via the occipital window at the level of the foramen magnum. These parameters included the peak systolic velocity (Vs), mean velocity (Vm), end-diastolic velocity (Vd), pulsatility index (PI), and resistance index (RI) of the basilar artery (BA) and bilateral vertebral artery (VA); and (2) The patient was placed in a lying upright position. The operator used a 2M doppler probe above the head to detect the bilateral middle cerebral artery (MCA) and bilateral anterior cerebral artery (ACA) through the temporal window at the superior margin of the temporal bone. ACA and bilateral posterior cerebral artery (PCA) of Vs, Vm, Vd, PI, RI. During the detection process, the depth of observation was adjusted to facilitate observation of the cerebral hemodynamic state of the entire artery within the detectable range. The cerebral hemodynamic parameters of transcranial doppler were detected and recorded by the same method in all patients.

Construction of clinical prediction models

Univariate and multifactorial logistic regression analysis combination was employed to screen the patients for factors that differed between groups and to generate odds ratio and 95% confidence interval (CI). The resulting predictive models were constructed and visualized in the form of column-line plots. A variance inflation factor was employed to assess the presence of multicollinearity among the variables. The area under the curve (AUC) for the subjects’ job characteristics was established to evaluate the variability of model predictions. The degree of correction of the model was evaluated through the utilization of decision curve analysis (DCA) and Hosmer-Lemeshow Goodness-of-Fit test; decision curves were plotted to illustrate the alteration in net return values in conjunction with shifts in risk probability. The optimal cut-off value of the prediction model was determined based on Jorden’s index, and the accuracy, sensitivity, and specificity were calculated.

Statistical analysis

SPSS 19.0 and R 3.6.2 software were used for processing and analyses, and measures conforming to normal distribution are expressed as mean ± SD; independent samples t-test was used for comparisons between groups; one-way analysis of variance was selected for differences between the three groups; Post-hoc tests were performed for indicators with significant differences, least significant difference test was used to test the indexes that were consistent with the homogeneity of variance, and Tamhane’s T2 test was used to test the indexes with uneven variance; Count data were expressed as percentages, and a χ² test was used between two or more groups constituting the ratio; variables with P < 0.05 in the one-way logistic regression analysis were included in the multifactor logistic regression model for correction, and variance inflation factor was calculated to determine covariance. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

RESULTS

Comparison of clinical data

There were no statistically significant differences among the three patient groups regarding age, gender, years of education, body mass index, or disease duration (P > 0.05). Detailed data are presented in Table 1.

Table 1 General information of patients in three groups.

Variables		SCZ group (n = 117)	PTSD group (n = 108)	RDD group (n = 105)	F	P value
Age (year), mean ± SD		26.97 ± 6.15	26.28 ± 5.49	27.10 ± 5.23	0.669	0.513
Gender	Male	54	52	35	5.645	0.06
	Female	63	56	70
Years of education		12.37 ± 3.32	12.49 ± 2.68	12.03 ± 2.86	0.688	0.503
BMI		23.13 ± 3.43	22.64 ± 3.00	22.94 ± 3.15	0.665	0.515
Course of disease (months)		4.81 ± 1.22	5.02 ± 1.08	4.95 ± 1.41	0.819	0.442

SCZ group: Schizophrenia with auditory verbal hallucinations; PTSD group: Post-traumatic stress disorder with auditory verbal hallucinations; RDD group: Recurrent depressive disorder with auditory verbal hallucinations; BMI: Body mass index.

Comparison of cerebral hemodynamic indexes among the three patient groups

The differences between the three patient groups in the indexes of Vm-MCA, Vd-MCA, Vm-ACA, Vm-PCA, Vs-PCA, Vd-PCA, Vm-BA, Vs-BA, Vd-BA, PI-BA, RI-BA, Vm-VA, Vd-VA, PI-VA and RI-VA were statistically significant (P < 0.05), and the specific analysis results are shown in Table 2.

Table 2 Cerebral hemodynamic indices in three groups of patients, mean ± SD.

Variables	SCZ group (n = 117)	PTSD group (n = 108)	RDD group (n = 105)	F	P value
Vm-MCA	57.25 ± 10.26	56.21 ± 8.15	59.36 ± 9.16b	3.214	0.041
Vs-MCA	88.52 ± 15.03	87.25 ± 13.64	91.26 ± 13.26b	2.260	0.106
Vd-MCA	41.05 ± 11.63	40.55 ± 10.69	44.69 ± 8.19ab	5.143	0.006
PI-MCA	0.81 ± 0.18	0.82 ± 0.15	0.79 ± 0.21	1.255	0.287
RI-MCA	0.55 ± 0.19	0.57 ± 0.19	0.55 ± 0.15	0.777	0.461
Vm-ACA	48.75 ± 10.82	47.65 ± 9.95	44.25 ± 10.21a,b	5.597	0.004
Vs-ACA	74.36 ± 12.09	73.69 ± 10.21	71.15 ± 12.95a	2.220	0.110
Vd-ACA	31.95 ± 9.11	32.66 ± 8.81	34.21 ± 9.47	1.764	0.173
PI-ACA	0.86 ± 0.21	0.86 ± 0.22	0.79 ± 0.32	2.043	0.131
RI-ACA	0.58 ± 0.21	0.55 ± 0.14	0.59 ± 0.15	1.764	0.173
Vm-PCA	34.21 ± 7.67	37.65 ± 6.35a	33.32 ± 5.54b	12.843	< 0.001
Vs-PCA	51.06 ± 8.15	54.84 ± 8.69a	50.94 ± 7.47b	8.102	< 0.001
Vd-PCA	22.61 ± 6.95	24.36 ± 5.58	21.33 ± 5.49b	6.699	0.001
PI-PCA	0.80 ± 0.19	0.77 ± 0.14	0.82 ± 0.15b	2.651	0.072
RI-PCA	0.59 ± 0.18	0.63 ± 0.18	0.63 ± 0.15	1.824	0.163
Vm-BA	35.61 ± 7.64	38.21 ± 6.65a	35.24 ± 7.19b	5.506	0.004
Vs-BA	55.26 ± 10.62	57.95 ± 8.65a	54.86 ± 9.16b	3.339	0.037
Vd-BA	22.82 ± 7.19	24.69 ± 6.17a	22.17 ± 7.71b	3.703	0.026
PI-BA	0.81 ± 0.21	0.76 ± 0.15	0.83 ± 0.19b	4.323	0.014
RI-BA	0.56 ± 0.12	0.52 ± 0.15	0.58 ± 0.12b	5.605	0.004
Vm-VA	35.61 ± 7.65	38.19 ± 6.65	40.22 ± 7.49a,b	11.198	< 0.001
Vs-VA	56.85 ± 12.43	58.41 ± 10.65	59.17 ± 9.33	1.313	0.270
Vd-VA	18.21 ± 7.45	24.16 ± 7.95a	23.27 ± 6.13a,b	22.261	< 0.001
PI-VA	0.78 ± 0.17	0.73 ± 0.15a	0.73 ± 0.14a	4.129	0.017
RI-VA	0.58 ± 0.15	0.52 ± 0.13a	0.54 ± 0.15	3.444	0.033

^aP < 0.05, compared to schizophrenia with auditory verbal hallucinations group.

^bP < 0.05, compared to PTSD group.

SCZ group: Schizophrenia with auditory verbal hallucinations; PTSD group: Post-traumatic stress disorder with auditory verbal hallucinations; RDD group: Recurrent depressive disorder with auditory verbal hallucinations; Vm: Mean velocity; Vs: Peak systolic velocity; Vd: End-diastolic velocity; PI: Pulsatility index; RI: Resistance index; MCA: Middle cerebral artery; ACA: Anterior cerebral artery; PCA: Posterior cerebral artery; BA: Basilar artery; VA: Vertebral artery.

Logistic regression analysis of patients with SCZ and AVHs

We performed univariate logistic regression analysis of patient-related hemodynamic indexes, with SCZ with AVHs as 1, non-SCZ with AVHs as 0, and the original values of hemodynamic indices entered. The results showed that Vd-MCA, Vm-ACA, Vm-VA, Vd-VA, and PI-VA were statistically different, and incorporating them into logistic multifactorial regression analyses showed that increasing Vm-ACA is an independent risk factor for SCZ with AVHs, and decreasing Vm-VA, Vd-MCA, and Vd-VA are independent protective factors for SCZ with AVHs. The difference was statistically significant (P < 0.05), and the detailed analysis results are shown in Table 3.

Table 3 Univariate and multivariate logistic regression analyses in patients with schizophrenia with auditory verbal hallucinations.

Variables	Univariate logistic regression analysis					Multivariate logistic regression analysis
	β	SE	Z	P value	OR (95%CI)	β	SE	Z	P value	OR (95%CI)
Vm-MCA	-0.02	0.01	-1.60	0.110	0.98 (0.95-1.01)	-	-	-	-	-
Vs-MCA	-0.01	0.01	-1.42	0.154	0.99 (0.97-1.01)	-	-	-	-	-
Vd-MCA	-0.04	0.01	-2.60	0.009	0.96 (0.94-0.99)	-0.06	0.02	-3.43	< 0.001	0.94 (0.91-0.97)
PI-MCA	0.51	0.68	0.75	0.455	1.66 (0.44-6.27)	-	-	-	-	-
RI-MCA	-0.33	0.77	-0.42	0.673	0.72 (0.16-3.27)	-	-	-	-	-
Vm-ACA	0.04	0.01	3.05	0.002	1.04 (1.01-1.07)	0.06	0.02	3.70	< 0.001	1.06 (1.03-1.10)
Vs-ACA	0.02	0.01	1.88	0.060	1.02 (1.00-1.04)	-	-	-	-	-
Vd-ACA	-0.03	0.01	-1.80	0.072	0.97 (0.95-1.00)	-	-	-	-	-
PI-ACA	0.85	0.51	1.65	0.099	2.33 (0.85-6.36)	-	-	-	-	-
RI-ACA	-0.18	0.75	-0.24	0.812	0.84 (0.19-3.65)	-	-	-	-	-
Vm-PCA	0.02	0.02	0.99	0.321	1.02 (0.98-1.06)	-	-	-	-	-
Vs-PCA	0.00	0.02	0.11	0.909	1.00 (0.97-1.04)	-	-	-	-	-
Vd-PCA	0.03	0.02	1.49	0.136	1.03 (0.99-1.08)	-	-	-	-	-
PI-PCA	-1.00	0.81	-1.24	0.214	0.37 (0.08-1.78)	-	-	-	-	-
RI-PCA	-1.53	0.85	-1.81	0.071	0.22 (0.04-1.14)	-	-	-	-	-
Vm-BA	0.01	0.02	0.36	0.717	1.01 (0.97-1.04)	-	-	-	-	-
Vs-BA	0.00	0.01	0.29	0.770	1.00 (0.98-1.03)	-	-	-	-	-
Vd-BA	0.01	0.02	0.65	0.514	1.01 (0.98-1.05)	-	-	-	-	-
PI-BA	-0.75	0.68	-1.11	0.267	0.47 (0.13-1.78)	-	-	-	-	-
RI-BA	-2.04	1.18	-1.72	0.085	0.13 (0.01-1.33)	-	-	-	-	-
Vm-VA	-0.08	0.02	-4.16	< 0.001	0.92 (0.89-0.96)	-0.11	0.02	-4.72	< 0.001	0.90 (0.86-0.94)
Vs-VA	-0.02	0.01	-1.55	0.121	0.98 (0.96-1.01)	-	-	-	-	-
Vd-VA	-0.11	0.02	-4.85	< 0.001	0.90 (0.86-0.94)	-0.13	0.03	-5.03	< 0.001	0.88 (0.84-0.93)
PI-VA	2.11	0.89	2.38	0.017	8.26 (1.45-47.10)	-	-	-	-	-
RI-VA	1.30	0.90	1.45	0.148	3.66(0.63-21.25)	-	-	-	-	-

SE: Standard error of mean; CI: Confidence interval; OR: Odds ratio; Vm: Mean velocity; Vs: Peak systolic velocity; Vd: End-diastolic velocity; PI: Pulsatility index; RI: Resistance index; MCA: Middle cerebral artery; ACA: Anterior cerebral artery; PCA: Posterior cerebral artery; BA: Basilar artery; VA: Vertebral artery.

Construction and evaluation of a risk column chart for patients with SCZ with AVHs

Based on the logistic multifactorial regression analysis, we established the column line graph model using Vm-ACA, Vm-VA, Vd-MCA, and Vd-VA as independent factors. Each factor corresponded to a score in the column line graph, and the greater the sum of scores of all the factors of the patient, the greater the probability of the AVH patient having SCZ (Figure 2A). We plotted the receiver operating characteristic (ROC) curve to evaluate the predictive ability of the column-line graphical model (Figure 2B), and the AUC of this regression equation was 0.82 (95%CI: 0.76-0.87), suggesting that high accuracy of the predictive model. The H-L test and calibration curve results showed that the predicted values were not significantly different from the measured values (Figure 2C), and the model was well fitted (P = 0.166 > 0.05). We constructed a DCA of the risk column-line diagram for patients with SCZ with AVHs to make the column-line diagram model clinically useful. This decision curve demonstrated comparable net benefits in the range of 0.2 to 1.0 based on a risk column-line diagram model of non-suicidal self-injury in adolescents with depression (Figure 2D). At the optimal critical value, the model had an accuracy of 0.78 (95%CI: 0.72-0.83), a sensitivity of 0.72 (95%CI: 0.64-0.81), a specificity of 0.83 (95%CI: 0.76-0.90), a positive predictive value (PPV) of 0.79 (95%CI: 0.71-0.87), a negative predictive value (NPV) of 0.77 (95%CI: 0.70-0.84), with an optimal cut-off value of 0.464 (Table 4).

Figure 2 Construction and evaluation of model 1. A: Nomogram; B: Receiver operating characteristic curve; C: Calibration curve; D: Decision curve analysis curves. MCA: Middle cerebral artery; VA: Vertebral artery; Vm: Mean velocity; Vd: End-diastolic velocity; AUC: Area under the curve; CI: Confidence interval.

Table 4 Confusion matrix analysis for clinical prediction model 1.

AUC (95%CI)	Accuracy (95%CI)	Sensitivity (95%CI)	Specificity (95%CI)	PPV (95%CI)	NPV (95%CI)	Cut off
0.82 (0.76-0.87)	0.78 (0.72-0.83)	0.72 (0.64-0.81)	0.83 (0.76-0.90)	0.79 (0.71-0.87)	0.77 (0.70-0.84)	0.464

AUC: Area under the curve; PPV: Positive predictive value; NPV: Negative predictive value; CI: Confidence interval.

Logistic regression analysis of patients with PTSD with AVHs

Univariate logistic regression analysis of patient-related hemodynamic indexes was performed, with the original values of cerebral hemodynamic indexes entered as 1 for patients with PTSD with AVHs, and 0 for patients without PTSD with AVHs. The Vm-MCA, Vs-MCA, Vd-MCA, Vm-ACA, Vm-PCA, Vs-PCA, Vd-PCA, PI-PCA, Vm-BA, Vs-BA, and RI-BA were statistically different for the eleven indices. The results of including them in logistic multifactorial regression analyses showed that rising Vm-ACA, Vm-PCA, Vs-PCA, Vd-PCA, Vm-BA, and Vs-BA are independent risk factors for PTSD with AVHs, and falling Vm-MCA, Vs-MCA, Vd-MCA, PI-PCA, and RI-BA are independent protective factors for PTSD with AVHs. The differences were statistically significant (P < 0.05), and the detailed analyses are shown in Table 5.

Table 5 Univariate and multivariate logistic regression analyses in patients with post-traumatic stress disorder with auditory verbal hallucinations.

Variables	Univariate logistic regression analysis					Multivariate logistic regression analysis
	β	SE	Z	P	OR (95%CI)	β	SE	Z	P	OR (95%CI)
Vm-MCA	-0.04	0.02	-2.59	0.010	0.96 (0.93-0.99)	-0.06	0.02	-2.74	0.006	0.95 (0.91-0.98)
Vs-MCA	-0.02	0.01	-2.14	0.033	0.98 (0.96-0.99)	-0.04	0.01	-2.75	0.006	0.96 (0.94-0.99)
Vd-MCA	-0.05	0.02	-3.05	0.002	0.96 (0.93-0.98)	-0.06	0.02	-2.71	0.007	0.95 (0.91-0.98)
PI-MCA	1.20	0.76	1.58	0.114	3.33 (0.75-14.84)	-	-	-	-	-
RI-MCA	0.66	0.79	0.83	0.404	1.93 (0.41-9.11)	-	-	-	-	-
Vm-ACA	0.03	0.01	2.41	0.016	1.03 (1.01-1.06)	0.05	0.02	2.45	0.014	1.05 (1.01-1.09)
Vs-ACA	0.02	0.01	1.57	0.115	1.02 (1.00-1.04)	-	-	-	-	-
Vd-ACA	-0.02	0.02	-1.23	0.217	0.98 (0.95-1.01)	-	-	-	-	-
PI-ACA	0.81	0.51	1.60	0.110	2.24 (0.83-6.04)	-	-	-	-	-
RI-ACA	-1.90	0.96	-1.98	0.048	0.15 (0.02-0.98)	-	-	-	-	-
Vm-PCA	0.12	0.03	4.71	< 0.001	1.13 (1.08-1.19)	0.12	0.03	3.61	< 0.001	1.13 (1.06-1.20)
Vs-PCA	0.06	0.02	3.31	< 0.001	1.06 (1.03-1.10)	0.07	0.02	2.69	0.007	1.07 (1.02-1.12)
Vd-PCA	0.10	0.03	3.73	< 0.001	1.11 (1.05-1.17)	0.13	0.04	3.54	< 0.001	1.14 (1.06-1.23)
PI-PCA	-2.55	1.01	-2.52	0.012	0.08 (0.01-0.57)	-2.77	1.38	-2.01	0.044	0.06 (0.00-0.93)
RI-PCA	-0.20	0.82	-0.24	0.810	0.82 (0.16-4.12)	-	-	-	-	-
Vm-BA	0.06	0.02	3.01	0.003	1.06 (1.02-1.11)	0.10	0.03	3.50	< 0.001	1.10 (1.04-1.17)
Vs-BA	0.04	0.02	2.46	0.014	1.04 (1.01-1.07)	0.05	0.02	2.47	0.013	1.06 (1.01-1.10)
Vd-BA	0.05	0.02	2.57	0.010	1.05 (1.01-1.10)	-	-	-	-	-
PI-BA	-2.55	0.85	-3.00	0.003	0.08 (0.01-0.41)	-	-	-	-	-
RI-BA	-3.38	1.10	-3.07	0.002	0.03 (0.00-0.30)	-4.07	1.46	-2.79	0.005	0.02 (0.00-0.30)
Vm-VA	-0.04	0.02	-2.05	0.040	0.96 (0.92-0.99)	-	-	-	-	-
Vs-VA	-0.01	0.01	-0.56	0.574	0.99 (0.97-1.02)	-	-	-	-	-
Vd-VA	0.02	0.02	0.90	0.366	1.02 (0.98-1.06)	-	-	-	-	-
PI-VA	0.04	0.95	0.05	0.964	1.04 (0.16-6.71)	-	-	-	-	-
RI-VA	-1.06	1.00	-1.06	0.291	0.35 (0.05-2.48)	-	-	-	-	-

SE: Standard error of mean; CI: Confidence interval; OR: Odds ratio; Vm: Mean velocity; Vs: Peak systolic velocity; Vd: End-diastolic velocity; PI: Pulsatility index; RI: Resistance index; MCA: Middle cerebral artery; ACA: Anterior cerebral artery; PCA: Posterior cerebral artery; BA: Basilar artery; VA: Vertebral artery.

Construction and evaluation of a risk column chart for patients with PTSD with AVHs

Based on the results of logistic multifactor regression analysis, we established a column-line graph model with Vm-MCA, Vs-MCA, Vd-MCA, Vm-ACA, Vm-PCA, Vs-PCA, Vd-PCA, PI-PCA, Vm-BA, Vs-BA, and RI-BA as independent factors (Figure 3A). We plotted the ROC curve to evaluate the predictive ability of the column-line graphical model (Figure 3B). The AUC of this regression equation was 0.88 (95%CI: 0.83-0.92), suggesting that the accuracy of the predictive model of this ROC curve was high. After the H-L test and calibration curve results showed that the predicted values were not significantly different from the measured values (Figure 3C), and the model was well fitted (P = 0.983 > 0.05); The construction of decision curves showed that the net benefits were comparable, in the range of 0.2 to 1.0 according to the PTSD with AVHs column line plot model (Figure 3D). At the optimal threshold, the model had an accuracy of 0.80 (95%CI: 0.75-0.89), a sensitivity of 0.79 (95%CI: 0.71-0.86), a specificity of 0.79 (95%CI: 0.71-0.87), a PPV of 0.79 (95%CI: 0.71-0.87), an NPV of 0.82 (95%CI: 0.74-0.89), with an optimal cut-off value of 0.534 (Table 6).

Figure 3 Construction and evaluation of model 2. A: Nomogram; B: Receiver operating characteristic curve; C: Calibration curve; D: Decision curve analysis curves. MCA: Middle cerebral artery; ACA: Anterior cerebral artery; PCA: Posterior cerebral artery; BA: Basilar artery; Vm: Mean velocity; Vs: Peak systolic velocity; Vd: End-diastolic velocity; PI: Pulsatility index; RI: Resistance index; AUC: Area under the curve; CI: Confidence interval.

Table 6 Confusion matrix analysis for clinical prediction model 2.

AUC (95%CI)	Accuracy (95%CI)	Sensitivity (95%CI)	Specificity (95%CI)	PPV (95%CI)	NPV (95%CI)	Cut off
0.88 (0.83-0.92)	0.80 (0.75-0.89)	0.79 (0.71-0.86)	0.79 (0.71-0.87)	0.79 (0.71-0.87)	0.82 (0.74-0.89)	0.534

AUC: Area under the curve; PPV: Positive predictive value; NPV: Negative predictive value; CI: Confidence interval.

Logistic regression analysis of patients with RDD with AVHs

Univariate logistic regression analysis of patient-related cerebral hemodynamic indices were performed, with patients with RDD with AVHs as 1 and non-RDD with AVHs as 0. We entered the original values of the cerebral hemodynamic indexes. The results identified statistically significant differences in Vm-MCA, Vs-MCA, Vd-MCA, Vm-ACA, Vs-ACA, PI-ACA, Vm-PCA, Vs-PCA, Vd-PCA, PI-PCA, PI-BA, RI-BA, Vm-VA, and Vd-VA. We included these factors in logistic regression analyses and found that Vm-MCA, Vd-MCA, RI-BA, Vm-VA, and Vd-VA were independent risk factors, and Vm-ACA, Vs-ACA, Vm-PCA, Vs-PCA, and Vd-PCA were independent protective factors. The difference was statistically significant and detailed analysis is shown in Table 7.

Table 7 Univariate and multivariate logistic regression analyses in patients with recurrent depressive disorder with auditory verbal hallucinations.

Variables	Univariate logistic regression analysis					Multivariate logistic regression analysis
	β	SE	Z	P	OR (95%CI)	β	SE	Z	P	OR (95%CI)
Vm-MCA	0.03	0.01	2.35	0.019	1.03 (1.01-1.06)	0.05	0.02	3.12	0.002	1.05 (1.02-1.08)
Vs-MCA	0.02	0.01	2.00	0.045	1.02 (1.01-1.03)	-	-	-	-	-
Vd-MCA	0.04	0.01	3.11	0.002	1.04 (1.01-1.06)	0.06	0.01	3.91	< 0.001	1.06 (1.03-1.09)
PI-MCA	-0.88	0.65	-1.36	0.174	0.41 (0.12-1.48)	-	-	-	-	-
RI-MCA	-0.13	0.65	-0.21	0.837	0.87 (0.24-3.13)	-	-	-	-	-
Vm-ACA	-0.04	0.01	-3.16	0.002	0.96 (0.94-0.99)	-0.05	0.01	-3.45	< 0.001	0.95 (0.93-0.98)
Vs-ACA	-0.02	0.01	-2.05	0.041	0.98 (0.96-0.99)	-0.03	0.01	-2.14	0.032	0.97 (0.95-0.99)
Vd-ACA	0.02	0.01	1.77	0.076	1.02 (1.00-1.05)	-	-	-	-	-
PI-ACA	-0.96	0.48	-2.00	0.045	0.38 (0.15-0.98)	-	-	-	-	-
RI-ACA	0.78	0.70	1.11	0.269	2.17 (0.55-8.59)	-	-	-	-	-
Vm-PCA	-0.06	0.02	-3.10	0.002	0.94 (0.91-0.98)	-0.06	0.02	-2.93	0.003	0.94 (0.90-0.98)
Vs-PCA	-0.03	0.01	-1.96	0.050	0.97 (0.94-0.99)	-0.04	0.02	-2.09	0.036	0.96 (0.93-0.99)
Vd-PCA	-0.06	0.02	-2.86	0.004	0.94 (0.91-0.98)	-0.07	0.02	-2.92	0.004	0.93 (0.89-0.98)
PI-PCA	1.54	0.76	2.04	0.041	4.67 (1.06-20.51)	-	-	-	-	-
RI-PCA	0.81	0.70	1.16	0.247	2.26 (0.57-8.95)	-	-	-	-	-
Vm-BA	-0.03	0.02	-1.87	0.062	0.97 (0.94-1.00)	-	-	-	-	-
Vs-BA	-0.02	0.01	-1.48	0.140	0.98 (0.96-1.01)	-	-	-	-	-
Vd-BA	-0.03	0.02	-1.84	0.066	0.97 (0.94-1.00)	-	-	-	-	-
PI-BA	1.49	0.65	2.30	0.021	4.44 (1.25-15.80)	-	-	-	-	-
RI-BA	2.65	0.97	2.74	0.006	14.10 (2.12-93.73)	3.23	1.16	2.79	0.005	25.30 (2.62-244.31)
Vm-VA	0.06	0.02	3.71	< .001	1.07 (1.03-1.10)	0.09	0.02	4.39	< 0.001	1.09 (1.05-1.13)
Vs-VA	0.01	0.01	1.22	0.223	1.01 (0.99-1.04)	-	-	-	-	-
Vd-VA	0.04	0.02	2.41	0.016	1.04 (1.01-1.07)	0.07	0.02	3.61	< 0.001	1.07 (1.03-1.11)
PI-VA	-1.15	0.78	-1.48	0.140	0.32 (0.07-1.46)	-	-	-	-	-
RI-VA	-0.28	0.82	-0.35	0.729	0.75 (0.15-3.74)	-	-	-	-	-

SE: Standard error of mean; CI: Confidence interval; OR: Odds ratio; Vm: Mean velocity; Vs: Peak systolic velocity; Vd: End-diastolic velocity; PI: Pulsatility index; RI: Resistance index; MCA: Middle cerebral artery; ACA: Anterior cerebral artery; PCA: Posterior cerebral artery; BA: Basilar artery; VA: Vertebral artery.

Construction and evaluation of a risk column chart for patients with RDD with AVHs

We established a column-linear graphical model for predicting RDD with AVHs based on logistic multifactor regression analysis, with Vm-MCA, Vd-MCA, RI-BA, Vm-VA, Vd-VA, Vm-ACA, Vs-ACA, Vm-PCA, Vs-PCA, and Vd-PCA as independent factors (Figure 4A). We plotted the ROC curve to evaluate the predictive ability of the column-line graphical model (Figure 4B), and the AUC of this regression equation was 0.81 (95%CI: 0.77-0.86), suggesting that the accuracy of the predictive model of this ROC curve was high. After the H-L test, the goodness-of-fit test and calibration curve results showed that the predicted values were not significantly different from the measured values (Figure 4C), and the model was well fitted (P = 0.140 > 0.05); The DCA for constructing a column-line diagram of RDD with AVHs showed that the net benefits were comparable in the range of 0.1 to 1.0 (Figure 4D). At the optimal critical value, the model had an accuracy of 0.75 (95%CI: 0.70-0.80), a sensitivity of 0.74 (95%CI: 0.68-0.80), a specificity of 0.78 (95%CI: 0.70-0.86), a PPV of 0.88 (95%CI: 0.83-0.92), an NPV of 0.58 (95%CI: 0.50-0.66), with an optimal cut-off value of 0.314 (Table 8).

Figure 4 Construction and evaluation of model 3. A: Nomogram; B: Receiver operating characteristic curve; C: Calibration curve; D: Decision curve analysis curves. MCA: Middle cerebral artery; ACA: Anterior cerebral artery; PCA: Posterior cerebral artery; BA: Basilar artery; VA: Vertebral artery; Vm: Mean velocity; Vs: Peak systolic velocity; Vd: End-diastolic velocity; AUC: Area under the curve; CI: Confidence interval.

Table 8 Confusion matrix analysis for clinical prediction model 3.

AUC (95%CI)	Accuracy (95%CI)	Sensitivity (95%CI)	Specificity (95%CI)	PPV (95%CI)	NPV (95%CI)	Cut off
0.81 (0.77-0.86)	0.75 (0.70-0.80)	0.74 (0.68-0.80)	0.78 (0.70-0.86)	0.88 (0.83-0.92)	0.58 (0.50-0.66)	0.314

AUC: Area under the curve; PPV: Positive predictive value; NPV: Negative predictive value; CI: Confidence interval.

DISCUSSION

AVHs can occur in patients with various neurological and psychiatric disorders and in the general population. They are currently diagnosed using subjective approaches, which can lead to increased rates of misdiagnosis and mistreatment of disease. Misdiagnosis and mistreatment or inappropriate treatment strategies can lead to unnecessary treatment side effects and mental suffering, and even some degree of cognitive impairment and structural brain damage. The brain is devoid of energy storage, requiring energy to be transported through cerebral perfusion[12]. Consequently, irregularities in cerebral hemodynamics signify alterations in neuronal metabolism and activity. These physiological and pathological modifications of the brain occur well before any changes in brain anatomy[13].

Our study shows that increased Vm-ACA is a risk factor for SCZ with AVHs, while Vd-MCA, Vm-VA, and Vd-VA are independent protective factors. SCZ is frequently accompanied by alterations in cerebral blood flow, and auditory hallucinations are an auditory processing symptom that involves the cerebral cortex, particularly the auditory cortex and associated connection areas[14,15]. The ACA provides blood supply to the frontal, medial and partial parietal lobes of the brain, regions involved in a number of important functions, including cognitive control, emotional regulation and language processing[16]. The VA provide blood supply to the brain stem, which is responsible for fundamental life-support functions, including primary auditory information processing, and the cerebellum and portions of the posterior brain[17]. Furthermore, SCZ accompanied by auditory hallucinations is frequently associated with neurotransmitter imbalances in the brain. In particular, dopamine overactivity has been linked to auditory hallucinations[18]. Neurotransmitter imbalances may also impact the regulatory function of cerebral blood vessels, leading to a localized deficiency in blood supply, particularly in the frontal lobe and brain stem regions. Further studies have shown that the connectivity between different areas of the brain, such as the frontal lobe, auditory cortex, hippocampus, and basal ganglia, is impaired, and that either reduced or abnormally increased blood flow to these areas leads to symptoms of verbal hallucinations in SCZ[19]. Similarly, as a consequence of functional abnormalities in the auditory cortex, prefrontal cortex and other aforementioned regions, local metabolic demands increase, causing blood vessels to dilate to enhance blood flow. However, due to malfunctions in blood flow regulation, this compensatory diastole may manifest itself as a rising fall in flow velocity at end-diastole, which in turn triggers end-diastolic flow velocity changes in the middle cerebral and VA. Furthermore, previous studies have suggested that individuals with SCZ may have altered sympathetic nervous system function. The heightened role of the sympathetic nervous system may be attributed to augmented vasodilation, in turn gives altering flow velocity.

The etiology of AVHs in individuals with PTSD may be associated with brain dysfunction resulting from psychological trauma. PTSD frequently results in elevated emotional and cognitive activation in patients, impacting regions of the brain responsible for information processing and emotional regulation, including the prefrontal lobe and amygdala[20-22]. The long-term recollection of stressful or traumatic experiences results in the brain becoming hyper-responsive to external stimuli, which can manifest as hallucinations, including auditory hallucinations of speech. These voices are often associated with traumatic experiences and may reflect the patient’s ongoing anxiety and fear of past traumatic events. Our study demonstrates that hemodynamic correlates of the PCA, the BA and the VA are significantly altered and are independent risk factors for PTSD patients with auditory hallucinations. Firstly, patients with PTSD frequently present with autonomic nervous system dysfunction, particularly sympathetic nervous system overactivation. This increased excitability results in vasoconstriction, which in turn leads to increased blood flow velocity[23]. The sustained stress and anxiety following a traumatic event may result in sustained contractions, increased sensitivity, and even the development of vasospasm. Secondly, when the brain is in a state of heightened emotional and cognitive activity, the brain requires an increased oxygen and nutritional supply, resulting in corresponding adaptive changes in brain blood flow. In addition, many studies have shown that people with PTSD often have problems with the cardiovascular system, such as high blood pressure. This disease also leads to increased vascular resistance, and consistent with the findings of this study, reduced RI in patients is an independent protective factor for PTSD with auditory hallucinations[24,25]. We found the increased peak systolic flow rate, mean flow rate, and end-diastolic flow rate of the MCA, in addition to the augmentation of the posterior artery pulse index and basal artery RI, were independent risk factors for depression with auditory hallucinations. Firstly, patients presenting with depression and concomitant auditory hallucinations and other psychiatric symptoms typically exhibit autonomic nervous system dysfunction, characterized by sympathetic nervous system activation. This may result in related cerebral hemodynamic alterations driven by increased blood vessel tension and contraction[26]. Secondly, patients with depression also exhibit imbalances in neurotransmitter function, predominantly manifesting as abnormalities in key neurotransmitters such as norepinephrine and serotonin[27]. Depression is typically associated with reduced norepinephrine levels. However, in individuals with depression and auditory hallucinations, there is a potential for increased sympathetic nerve activity, which may result in vasoconstriction. Serotonin deficiency is associated with blood vessel regulation, resulting in blood vessel constriction dysfunction and increased flow rate[28]. A previous study demonstrated that individuals with prolonged chronic depression may exhibit minor structural and functional alterations in the brain’s blood vessels, particularly in regions linked to emotional regulation, such as the prefrontal cortex and the cingulate gyrus. The enhanced blood flow necessary for these regions results in an increase in artery velocity within the brain[29]. In addition, there may be a chronic low-grade inflammatory response in the brains of people with depression. It is evident that inflammatory factors exert a profound influence on a multitude of physiological processes, including those associated with depressive symptoms and auditory hallucinations. Moreover, their impact on the nervous system extends beyond these manifestations, as they can also modulate the contractile function of blood vessels. This can manifest as increased vascular permeability and vascular endothelial dysfunction, leading to enhanced vascular contraction and flow rate, as well as augmented vascular resistance[30].

In this study, we constructed risk prediction models for SCZ with auditory hallucinations, PTSD with auditory hallucinations and simple depression with auditory hallucinations using logistic regression analysis to identify predictors. Nomogram prediction models for patients with different types of AVHs were successfully established using the visualization processing of the nomogram to determine the risk type for patients with AVHs. This is important in implementation of the disease prevention concept of “early detection, early diagnosis and early treatment”. The AUC of the each model was above 0.80, indicating that the model exhibits a high degree of differentiation. Furthermore, the H-L test of the calibration curve of the three models indicates P > 0.05, indicating that the calibration curve is relatively fit with the ideal standard curve. This suggests that the prediction model we constructed has a high calibration ability. The actual values of the various types of AVHs were found to be consistent with the predicted values of the model. Subsequently, we evaluated the clinical utility of the prediction model using clinical prediction curve analysis. Within a certain threshold, the model demonstrated a higher net gain compared to no intervention for all patients or intervention for all patients. Therefore, the graph model we constructed is an effective tool for predicting the risk of disease type in patients with AVHs and has significant clinical application value.

This study also faces several limitations. First, as a retrospective study, there are inherent limitations in the study design, which cannot completely avoid selection bias and information bias. Moreover, it was not possible to control all variables during data collection, which may have resulted in some potential influences not being adequately considered. Second, this study was a single-center study with all data coming from the same medical center, which may limit the generalizability and extrapolation of the findings. In addition, although a large sample size of patients was included in this study, the differences in prevalence and consultation for different disease types resulted in relatively small sample sizes for the subgroups, which may have had some impact on the validity of the statistical analyses. Inadequate sample sizes may lead to instability in the results, possibly affecting the accurate identification of risk factors for each type of AVHs. Finally, the optimal time to test patient cerebral hemodynamic parameters is not clear. Cerebral hemodynamic parameters can fluctuate over time due to several factors, such as stress, drug effects, and circadian rhythms. Repeated measurements were not performed in this study to capture temporal variations, which may result in an incomplete and inaccurate assessment of patients’ cerebral hemodynamic characteristics. Based on the above limitations, the following recommendations are made for future studies: Conduct a multicenter, large-sample prospective study to broadly include patients from different regions and ethnicities to increase the diversity and representativeness of the sample. At the same time, the sample size of each subgroup should be expanded to enhance the efficacy of statistical tests.

CONCLUSION

In the study design, the clinical characteristics of patients are recorded in detail, and the effects of confounding factors are controlled through stratified analyses and multivariate regression. Longitudinal studies are needed to measure cerebral hemodynamic indexes at multiple time points and in different physiological and pathological states of patients to fully consider the influence of time factors on the results and to more comprehensively understand the changing law of cerebral hemodynamics.