Yoshio Hirayasu, MD, PhD, Department of Psychiatry, Graduate School of Medicine, Yokohama City University, 3-9 Fukuura, Kanazawa-ku, Yokohama, Kanagawa 236-0004, Japan. Email: email@example.com
Yoshio Hirayasu, MD, PhD, Department of Psychiatry, Graduate School of Medicine, Yokohama City University, 3-9 Fukuura, Kanazawa-ku, Yokohama, Kanagawa 236-0004, Japan. Email: firstname.lastname@example.org
Aim: Although recent studies suggest abnormalities of the cerebral cortex, limbic structures, and brain stem regions in panic disorder (PD), the extent to which the midbrain is associated with PD pathophysiology is unclear. The aim of this study was to investigate structural abnormalities of the midbrain using magnetic resonance imaging and to determine if there is a clinical correlation between midbrain volume and clinical measurements in patients with PD.
Methods: Thirty-eight patients with PD (PD group) and 38 healthy controls (HC group) participated in this study. The midbrain was measured with a manual tracing method with high spatial resolution magnetic resonance imaging. The Panic Disorder Severity Scale and Global Assessment of Functioning were used to examine the correlation between volume abnormality and clinical symptoms and functioning in the PD group.
Results: Relative midbrain volume was larger in the PD group than in the HC group. The relative volume of the dorsal midbrain was larger in the PD group, while the volume of the ventral midbrain was not. We found a significant positive correlation between relative dorsal midbrain volume and total Panic Disorder Severity Scale score, and a significant negative correlation between relative dorsal midbrain volume and Global Assessment of Functioning score in the PD group.
Conclusions: Our findings suggest that the dorsal midbrain is associated with PD pathophysiology. The midbrain volume increase may reflect PD severity.
AN ESSENTIAL FEATURE of panic disorder (PD), a type of anxiety disorder, is the recurrence of spontaneous and incapacitating panic attacks that are not due to any specific situation or environmental factor. A panic attack is characterized by a discrete period of intense fear or discomfort that develops abruptly and reaches a peak within 10 minutes. The symptoms of a panic attack include a fear of losing control or dying, and various physical symptoms, such as heart palpitations, a smothering sensation, dizziness, nausea, or sweating.
Gorman et al. proposed a neuroanatomical model specific to PD, in which unexpected panic attacks involve brain stem nuclei. They hypothesized that selective serotonin reuptake inhibitors (SSRI) work through stabilization of brain stem nuclei.1 Viscerosensory information is conveyed to the amygdala, which coordinates autonomic and behavioral responses.2 Efferent pathways of the amygdala include the parabrachial nucleus,3 the lateral4 and paraventricular5 nucleus of the hypothalamus, and the locus ceruleus.6 A projection to the periaqueductal gray (PAG) region is responsible for defensive behaviors and postural freezing, which may be the animal equivalent of phobic avoidance.7 Coplan and Lydiard reviewed fear circuitries and modulation of neurotransmitter systems in PD. They concluded that the serotonin (5-hydroxytryptamine [5-HT]) system, which includes the dorsal (DRN) and median raphe nuclei (MRN) in the brainstem, is relevant to panic neurocircuits.8 Electrical or chemical stimulation of the PAG evokes escape, a defensive behavior that has been related to panic attacks. Evidence in the literature suggests that changes in 5-HT-mediated neurotransmission at the level of the PAG are involved in the pathophysiology of panic disorder and in the mode of action of panicolytic drugs. Anxiety in PD may result from either enhanced or diminished 5-HT neurotransmission.9,10
Several functional magnetic resonance imaging (fMRI) studies have reported abnormal function in the brainstem of patients with PD. Boshuisen et al. reported hyperactivity in the midbrain of patients with PD compared to control participants,11 and Reiman et al. showed that lactate-induced panic attacks activate the midbrain.12 In a positron emission tomography (PET) study, Sakai et al. reported that glucose metabolism increases in the midbrain of untreated patients with PD.13 Previous structural neuroimaging studies in PD patients have shown abnormalities in the amygdala,14 parahippocampus,15 anterior cingulate cortex,16 and frontal regions.17 These previous findings support Coplan's and Gorman's hypotheses on PD. In the brainstem, only two structural imaging studies in patients with PD have reported midbrain volume changes by voxel-based morphometry (VBM).18,19 No study has quantitatively examined midbrain volume abnormalities in patients with PD using manual tracing of the region of interest (ROI). We hypothesize that the structure in the dorsal section of the midbrain has changed in PD. The aim of this study was to investigate structural midbrain abnormalities, mainly the dorsal section containing PAG, in patients with PD using a manual method and to determine the clinical correlation between midbrain volume and clinical PD measurements.
Thirty-eight patients with PD (PD group; 25 female, 13 male) and 38 healthy controls (HC group) participated in this study. We recruited inpatients and outpatients at Yokohama City University Hospital between March 2005 and August 2009. Healthy controls were recruited from the community and hospital staff between March 2006 and June 2008. All participants met the following criteria: age 19–57 years, IQ > 75, right-handed, no history of epilepsy, head trauma with loss of consciousness, neurological disorders, or substance abuse. Six patients had a family history of psychiatric disorders (one with panic disorder and five with depressive disorder). The Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I20) was used to diagnose panic disorder and rule out other past or current Axis I disorders. The Panic Disorder Severity Scale (PDSS21) and Global Assessment of Functioning (GAF) were used to identify patients with comorbid psychiatric disorders, who were excluded from the study except for patients with major depressive disorder, which is often associated with PD. Therefore, this study included six patients with a history of major depressive disorder and 22 patients with a history of agoraphobia, 11 of whom were suffering from agoraphobia at the time of the study.
Patients were receiving SSRI alone (n = 3); benzodiazepines alone (n = 7); SSRI and benzodiazepines (n = 22); serotonin-norepinephrine reuptake inhibitors (SNRI) and benzodiazepines (n = 2); SSRI, SNRI, and benzodiazepines (n = 1); tricyclic antidepressants and benzodiazepines (n = 2); and tetracyclic antidepressants, SSRI, and benzodiazepines (n = 1). Healthy controls were screened for Axis I disorders using the SCID (Non-patient Edition)20 and the Mini-International Neuropsychiatric Interview.22 None of the healthy controls had ever received any psychiatric treatment and none of their first-degree relatives had ever had Axis I disorders. The socioeconomic status (SES) of all participants and their parents were assessed with the Hollingshead Two-Factor Index.23 This study was approved by the Medical Research Ethics Committee of Yokohama City University. After providing a complete description of the study, we obtained written informed consent from all participants. Twenty panic disorder patients and 16 healthy controls in this study overlap with participants in our previous MRI study in PD.
MRI scans were acquired with a 1.5-T Magnetom Symphony system (Siemens Medical System, Erlangen, Germany) at Yokohama City University Hospital. A series of 128 contiguous T1-weighted slices in sagittal images was acquired using a Turbo FLASH sequence with the following parameters: echo time = 3.93 ms, repetition time = 1960 ms, inversion time = 1100 ms, flip angle = 15, field of view = 24 cm, matrix = 256 × 256 × 128, and voxel dimensions = 0.9375 × 0.9375 × 1.5 mm.
For measuring intracranial contents (ICC), we obtained 60 contiguous Turbo SE 3-mm axial slices throughout the brain with the following imaging parameters: TE = 93 ms, TR = 3400 ms, field of view = 24 cm, matrix = 256 × 256, and voxel dimensions = 0.9375 × 0.9375 × 3.0 mm. No gross abnormalities were found in the scans upon evaluation by a clinical neuroradiologist.
Regions of interest
The ROI was the midbrain, which was delineated manually using a medical image analysis software package (3D slicer, http://www.slicer.org/). We realigned the images using the line between the inferior border of the mammillary body and the posterior commissure21 and the midsagittal plane to correct head tilt, and resampled them into isotropic voxels (0.9375 × 0.9375 × 0.9375 mm3). Manual drawings of ROI were performed on the realigned and resampled axial slices.
The most superior plane of the midbrain was defined by an axial slice containing the inferior border of the mammillary body and posterior commissure.23 The most inferior plane of the midbrain was defined by an axial slice containing the lower border of the inferior colliculus (Fig. 1a). In the upper level of the midbrain, the ventral boundary between the midbrain and the diencephalon was defined on the most superior axial slice, in which the outside rims of the midbrain are clearly visible (Fig. 1b). We used the same ventral boundary between the midbrain and the diencephalon for all upper slices other than the most superior axial slice. On each axial slice, we traced inside the area surrounded by straight lines connecting the lateral rims of the midbrain and the lateral rims of the superior colliculi (Fig. 1c). In the medium level of the midbrain, the midbrain boundary appeared clearly. The anterior rim of the midbrain was the most ventral side of the midbrain. The posterior rim of the midbrain was the most dorsal side of the corpora quadrigemina. In the lower level of the midbrain, the pons next to the midbrain was excluded on every sagittal slice. The shape of the pons was smooth and ovoid (Fig. 1a,d).24,25 The cerebral aqueduct was excluded on all axial slices.
Additionally, we divided the midbrain into ventral and dorsal sections. We defined the boundary between ventral and dorsal sections as a straight line through the sulcus of bilateral cerebral peduncles on the most inferior axial slice of the midbrain, and used the same straight line dividing ventral and dorsal sections on all axial slices of the midbrain (Fig. 1e,f). We used cerebral peduncles as the boundary between the ventral and dorsal sections to measure the region containing the PAG because PAG is in the dorsal section of the midbrain with the boundary of a straight line through the sulcus of bilateral cerebral peduncles.
A.F. was blinded to the diagnoses and performed manual tracing of the ROI for all participants. ICC volume was used to control for head size differences. Relative volume was computed with the following formula: relative volume = (ROI volume/ICC volume) × 100 (%). Interrater reliability for ROI was examined in 10 randomly assigned cases by three raters (A.F., F.H., H.N.) blind to the diagnoses. The intra-class correlation coefficients (Cronbach's alpha) were 0.93 for total midbrain, 0.91/0.90 for ventral/dorsal sections of the midbrain, and 0.99 for ICC. Cronbach's alpha for intrarater reliability evaluated in 10 randomly assigned cases for A.F. was 0.96 for total midbrain, 0.92/0.95 for ventral/dorsal sections of the midbrain, and 0.99 for ICC.
Statistical analysis was performed using spss version 11.0J for Windows (spss, Chicago, IL, USA). We used t-tests to assess group differences in age, sex, handedness, participant and parental SES, years of education, Wechsler Adult Intelligence Scale Revised (WAIS-R) IQ scores, GAF scores, and ICC. We used repeated measures anova with group (panic disorder/control) and sex (male/female) as between-subjects factors, and section (ventral/dorsal) as a within-subject factor. We examined the correlation between volume of ROI and PDSS and GAF scores in PD patients using two-tailed Pearson's correlation coefficients. Significance levels were set at P < 0.05.
Table 1 shows demographic and clinical characteristics of the PD and HC groups. The two groups did not differ significantly in age, sex, handedness, participant SES, and years of education. Parental SES and WAIS-R IQ score show difference at trend level, which may be attributed to involvement of healthy control subjects partly recruited from medical personnel. GAF scores were significantly lower in the PD group than in the HC group.
Table 1. Demographic and clinical characteristics of patients with panic disorder and healthy control participants
Panic disorder patients (n = 38)
Control subjects (n = 38)
t(d.f. = 74)
Age range for panic disorder and healthy control groups was 19–57 years and 23–57 years, respectively.
Handedness was evaluated using the Edinburgh inventory and right-handedness is above zero. Range of panic disorder and healthy control groups was 0.5–1.0 and 0.4–1.0, respectively.
SES was assessed using the Hollingshead scale. A lower score indicates higher status.
GAF, Global Assessment of Functioning; PDSS, Panic Disorder Severity Scale (0–28); SES, socioeconomic status; WAIS-R, Wechsler Adult Intelligence Scale Revised.
We performed anova for the relative volume of the midbrain between groups (PD/HC) with sections (ventral/dorsal) and sex (male/female). There was a significant main effect of group, which indicates that the relative volume of the midbrain was larger in the PD group compared to the HC group. There was a significant interaction of group × section (F = 4.027; d.f. = 1.74; P = 0.049). To further examine the significance of the group × section interaction, we performed anova by group and section in ventral and dorsal sections, separately. While there was no significant group difference in the ventral section (F = 0.580; d.f. = 1.74; P = 0.449), there was a significant difference in the dorsal section (F = 14.469; d.f. = 1.74; P < 0.001), which indicates that the relative volume in the dorsal section of the midbrain is larger in the PD group relative to the HC group (Fig. 2). There was no significant interaction of group × section × sex (F = 0.752; d.f. = 1.74; P = 0.389).
To investigate the influence of differences between the two groups in WAIS-R and socioeconomic status on the results, we performed ancova for the relative volume of the midbrain with groups (PD/HC) and sex (male/female) as between-subjects factor, sections (ventral/dorsal) as within-subject factor, and WAIS-R score and parental socioeconomic status as covariates. We have obtained the same results as in the anova with group, sex and sections, which suggests that the differences between the two groups in WAIS-R score and parental socioeconomic status were not a contributing factor to the obtained results.
We divided the PD group into two groups according to the cut-off score for PDSS to examine the effect of current symptoms severity of PD patients on the midbrain volume in this study. There were 14 non-symptomatic (under eight points in PDSS) patients and 24 symptomatic (over eight points) patients. We performed anova for the relative volume of the midbrain with groups (symptomatic/non-symptomatic), sections (ventral/dorsal), and sex (male/female). There was not a significant main effect of group and interaction of group × section. There was a significant interaction of group × section × sex (F = 4.556; d.f. = 1.36; P = 0.040). To further examine the significant interaction of the group × section × sex, we performed anova with group and sex in ventral and dorsal sections, respectively. While there was no significant interaction of group × sex in the dorsal section (F = 0.909; d.f. = 1.36; P = 0.347), there was a significant interaction of group × sex in the ventral section (F = 4.187; d.f. = 1.36; P = 0.049). These results suggest that the ventral midbrain volume was larger in the symptomatic group than in the non-symptomatic group in female PD patients.
Correlation between relative volume of the midbrain and psychological evaluations
First, we computed Pearson's correlation between midbrain relative volumes and psychological evaluation scores. There was a significant positive correlation between the relative midbrain volume and total PDSS score in the PD group (total: r = 0.518, P = 0.001, dorsal: r = 0.438, P = 0.006) (Fig. 3a). There was a significant negative correlation between relative midbrain volume and GAF score in the PD group (total: r = −0.432, P = 0.007, dorsal: r =−0.481, P = 0.002) (Fig. 3b). In the dorsal section, these two correlations were also significant on partial correlation analysis covariated with duration of illness and medication (PDSS: r = 0.5163, P = 0.001; GAF: r = −0.4148, P = 0.012). There was not significant correlation between relative volume of midbrain and total dose of antidepressants (r = 0.076, P =0.649), benzodiazepines (r = 0.106, P = 0.527), duration of illness or medication in patients with PD.
In this study, the PD group included six patients with a history of major depression. Previous reports have suggested that patients with depressive episodes have changes in brain volume.26 To examine the effect of comorbidity with major depressive disorder on relative midbrain volume in patients with PD, we performed anova after excluding the six patients with a history of major depressive disorder. We obtained significant volume differences between groups and interaction of the group × section (F = 4.065; d.f. =1.68; P = 0.048). We found significant volume differences only between groups in the dorsal section of the midbrain (F = 11.418; d.f. = 1.68; P = 0.001). There was a significant correlation between dorsal midbrain volume and psychological evaluation using the PDSS (r = 0.402, P = 0.020) and GAF (r = −0.462, P = 0.007).
The present study demonstrated dorsal midbrain volume increases in patients with PD compared to healthy controls, regardless of sex, using manual tracing of ROI. This result is consistent with previous findings of VBM studies that also showed midbrain volume increases.18,19 There was a significant difference between the groups, and significant correlation between the volume and psychological evaluation score only in the dorsal section. The correlation was also significant on partial correlation analysis covariated with dose of medications. These findings suggest that the dorsal section of the midbrain may be more strongly involved with PD pathophysiology, and suggest that a midbrain volume increase may reflect the global severity of clinical symptoms in patients with PD. Moreover, we show that the ventral midbrain volume is larger in the symptomatic group than in the non-symptomatic group in female PD patients only. These findings might reflect sex difference of pathophysiology of PD in the structure of the midbrain.
Anatomically, the PAG, which is a cell-dense region, surrounds the midbrain aqueduct in the dorsal midbrain. In this study, the PAG was included in the dorsal section of the midbrain. The PAG consists of longitudinal columns of afferent inputs and output neurons, and intrinsic interneurons. The PAG contains many neurons that project to the brainstem, hypothalamus, thalamus, and amygdala, regions that also input to the PAG.27
Animal studies using electrical stimulation28 and intracerebral drug injection29 into the PAG indicate a role for 5-HT in the regulation of defense. 5-HT enhanced in the PAG inhibits unconditioned defense reactions to proximal threats.30 This 5-HT system in PAG is essential to the pathophysiology of PD.31 Ascending tracts of the neural network in PD include 5-HT neurons in the DRN and MRN.32 The DRN have been shown to receive input from many nuclei in the limbic system and elsewhere,8 while these neurites project into the PAG, hypothalamus, and amygdala.2 Descending tracts of the neural network in PD projecting from the amygdala into parabrachial nuclei,4 hypothalamus,5,7 locus coeruleus,6 and PAG,33 control fear behavior and relate physiological changes of the body. These studies suggest the probability of more important involvement of the midbrain, including the DRN, MRN, and PAG.31,32 Electrical stimulation in or near the dorsal PAG in humans had also induced intense states of fear and/or terror and autonomic outflow suggestive of the ‘fight or flight’ response.34
In fMRI study, distal threat elicits activity in the prefrontal cortices in humans. Whereas threat becomes proximal and immediate, brain activity shifted from the prefrontal cortex to the PAG.35 These results indicate the PAG involvement in the process of reflexive responding (e.g. fight or flight) to the aversive stimuli and threat. Furthermore, higher cortical systems control behavior when the degree of threat is appraised as non-life-endangering and guides the organism to choose the most effective and resourceful strategy for instrumental avoidance. This defense system for threat can conduct the important role of PAG in the neurobiology of panic disorder. This study showed volume changes in the dorsal midbrain in which the PAG is included. The correlation between dorsal midbrain volume and psychological evaluation score in patients with PD indicates the possibility of PAG participation in PD symptom severity.
There were methodological limitations in this study. First, the sample size of participants in both groups was not large. Second, we do not distinguish PD from other anxiety disorders with situation-dependent panic attacks. No previous reports have examined midbrain volume in patients with other anxiety disorders. While there was a positive correlation between midbrain volume and PDSS score for patients with PD, we have not evaluated midbrain volume and clinical symptoms in other anxiety disorders. Therefore, we cannot conclude whether midbrain volume increase is specific to PD. Third, we have shown volume increase of dorsal midbrain volume in PD only with manual tracing volumetry, not the automatic measuring method. Findings shown by several studies with ROI manual tracing are different from those with automatic volumetry, which could be accounted for the influence of the rater's proficiency of measuring ROI. Although ROI are manually traced with high intra-class correlation coefficients (more than 0.90 for each ROI) in this study, further studies using atlas-based automatic volumetry are needed to clarify the involvement of the dorsal midbrain including PAG in the pathophysiology of PD.
In conclusion, this study used a manual tracing method on MRI to demonstrate dorsal midbrain volume increases in patients with PD. The volume change of the ROI is associated with the clinical severity of PD symptoms and GAF score. These findings suggest that dorsal midbrain volume increases may be associated with PD pathophysiology.
This study was supported by a grant from the Ministry of Health, Labor and Welfare (Kokoro 200400762A, 200500806A, 200632005A, B) of Japan.