Long‐term benefits of mindfulness on white matter tracts underlying the cortical midline structures in panic disorder: A 2‐year longitudinal study

Abstract

Aims

We aimed to examine the long‐term benefits of mindfulness‐based cognitive therapy (MBCT) on white matter plasticity in the cortical midline structures (CMS) for a period of 2 years in patients with panic disorder and the relationships between white matter changes in the CMS and severity of state and trait symptoms.

Methods

Seventy‐one participants were enrolled and underwent diffusion tensor imaging at baseline and after 2 years (26 who received MBCT as an adjunct to pharmacotherapy [MBCT+PT], 20 treated with pharmacotherapy alone [PT‐alone], and 25 healthy controls [HCs]). The severity of symptoms and fractional anisotropy (FA) in white matter regions underlying the CMS were assessed at baseline and 2‐year follow‐up.

Results

The MBCT+PT group showed better outcomes after 2 years than the PT‐alone group. The groups showed different FA changes: the MBCT+PT group showed decreased FA in the left anterior cingulate cortex (ACC); the PT‐alone group showed increased FA in the bilateral dorsomedial prefrontal cortex, posterior cingulate cortex (PCC), and precuneus. Decreased white matter FA in the ACC, PCC, and precuneus was associated with improvements in the severity of state and trait symptoms in patients with panic disorder.

Conclusion

Alleviation of excessive white matter connectivity in the CMS after MBCT leads to improvements in clinical symptoms and trait vulnerability in patients with panic disorder. Our study provides new evidence for the long‐term benefits of MBCT on white matter plasticity and its clinical applicability as a robust treatment for panic disorder.

Panic disorder is one of the most prevalent psychiatric disorders, with a chronic and recurrent course from onset. ¹ , ² Although serotonergic antidepressants are effective first‐line treatments for some patients, a substantial proportion of patients experience relapse. ³ , ⁴ Recently, a novel psychotherapeutic approach that involves mindfulness was developed to treat anxiety and depression. ⁵ Since its introduction, mindfulness‐based therapy has become an increasingly popular intervention with significant benefits for patients with anxiety and depression. ⁶ , ⁷ , ⁸

Mindfulness‐based cognitive therapy (MBCT) is underscored by several features that may be beneficial for treating panic disorder. While conventional cognitive behavioral therapy focuses on altering maladaptive cognition and behaviors that perpetuate panic symptoms, MBCT emphasizes training to focus on the present moment and adopting acceptance. ⁹ Patients with panic disorder, thereby, develop their ability to cope with symptoms and co‐occurring ruminations, worrying, and depressed mood by exploring anxiety‐related experiences in an open and accepting manner. We previously implemented MBCT as an adjuvant to pharmacotherapy for patients with panic disorder; the administration of 8‐week MBCT was effective in reducing anxiety and panic symptoms ¹⁰ , ¹¹ and led to long‐term positive clinical outcomes, such as prevention of relapse. ¹² Furthermore, MBCT significantly decreased anxiety sensitivity, a trait‐like risk factor in the development of panic disorder, ¹⁰ , ¹¹ which predicted the remission of panic disorder after 1 year. ¹⁰ Neuroimaging research has indicated that mindfulness is associated with structural and functional changes in several cortical regions, such as the prefrontal, insular, and cingulate cortices. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ Nevertheless, the neurobiological mechanisms underpinning the therapeutic effects of MBCT in panic disorder remain unclear.

Diffusion tensor imaging (DTI) is a brain imaging technique that enables in vivo characterization of white matter tracts. White matter myelination and remodeling constitute a fundamental mechanism that enhances neuroplastic changes in the adult brain. ¹⁷ Changes in fractional anisotropy (FA) that originate from alterations in myelin and axonal density serve as a useful proxy of white matter plasticity after MBCT. In previous DTI studies of healthy participants, 4 weeks of mindfulness training improved the integrity of white matter tracts connecting the anterior cingulate and surrounding medial cortices. ¹⁸ , ¹⁹ The cortical midline structures (CMS), consisting of the cingulate, medial prefrontal, and precuneus cortices, are putative neural correlates of self‐related mental operations involving mindfulness ²⁰ , ²¹ ; however, studies investigating the long‐term neuroplastic changes after MBCT in the white matter tracts within these regions in clinical populations are scarce.

We aimed to examine the white matter plasticity of the CMS 2 years after MBCT in patients with panic disorder receiving maintenance pharmacotherapy (MBCT+PT) compared to patients treated with pharmacotherapy alone (PT‐alone) and healthy controls (HCs). We hypothesized that patients who received 8 weeks of MBCT as an adjuvant to pharmacotherapy would exhibit a distinctive pattern of neuroplastic changes in the white matter underlying the CMS at the 2‐year follow‐up, compared to the PT‐alone group. Furthermore, we explored the relationship between white matter changes and severity of state and trait symptoms, assuming that longitudinal FA changes related to MBCT would be associated with the symptomatic improvement of panic disorder.

Methods

Study design and participants

Patients with panic disorder were recruited from the outpatient psychiatry clinic at CHA Bundang Medical Center in Seongnam, Republic of Korea, between December 2011 and September 2016. The inclusion criteria were new diagnosis of panic disorder based on the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM‐IV‐TR) Axis I Disorders ²² and treatment‐naïve status. Patients were excluded if they had a history of major psychiatric comorbidities, such as psychotic, substance use, bipolar, and/or major depressive disorders; neurological disorders; traumatic brain injury; intellectual disability (i.e., intelligence quotient <70); clinically significant or unstable medical illness; and formal concurrent psychotherapy. Finally, handedness was assessed in all participants using the Edinburgh Handedness Inventory, ²³ and left‐handed individuals were excluded. HCs were recruited from the local community through print and online advertisements. The same exclusion criteria were applied to HCs, along with excluding those with a first‐degree relative with a psychiatric disorder.

Participants underwent brain magnetic resonance imaging (MRI) twice, at an interval of 2 years (mean ± standard deviation [SD] = 24.8 ± 1.4 months; range: 22–26 months). The severity of state and trait symptoms was measured at each time point. Current state symptoms were assessed using the Panic Disorder Severity Scale (PDSS), ²⁴ Beck Depression Inventory‐II (BDI‐II), ²⁵ and Beck Anxiety Inventory (BAI). ²⁶ Trait symptoms of panic disorder were assessed using the neuroticism subscale from the Neuroticism‐Extraversion‐Openness Personality Inventory (NEO‐N) and Anxiety Sensitivity Index‐Revised (ASI‐R). ²⁷ After the baseline scan, all patients with panic disorder were offered pharmacotherapy and asked to participate in adjunctive MBCT. Treatment was started with selective serotonin reuptake inhibitors (i.e., paroxetine, escitalopram, and sertraline) and benzodiazepines (i.e., lorazepam, alprazolam, diazepam, and clonazepam) according to the 2008 Korean Medication Algorithm for panic disorder. ²⁸ The patients voluntarily decided to participate in MBCT after receiving a sufficient explanation from therapists who were blinded to the purpose and hypotheses of the study. The patients in the MBCT+PT group had 1.5–2 h group sessions every week for 8 weeks along with pharmacotherapy. Each treatment group consisted of 6–10 patients with a mix of study participants and non‐participants, and the patients' participation in this study was not disclosed to the therapists. Patients were excluded from the MBCT+PT group if they either missed two or more treatment sessions or showed non‐adherence to the program (i.e., less than 80% homework completion). During the maintenance period, researchers blinded to the group assignment contacted the patients monthly to monitor medication adherence, determined by at least an 80% administration rate of the prescribed medication and dosage for 1 month. The PDSS was additionally administered to assess interim treatment response after 8 weeks, 6 months, or 1 year from the initiation of treatment. Remission of panic disorder was evaluated according to Ballenger's criteria as a PDSS total score ≤3, with no individual item having a score of >1. ²⁹ In the final analysis, data from 46 patients who completed both baseline and 2‐year follow‐up scans and demonstrated acceptable treatment adherence to MBCT and pharmacotherapy (26 MBCT+PT and 20 PT‐alone patients) and 25 HCs were included.

All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The current study was reviewed and approved by the Institutional Review Board of CHA Bundang Medical Center. All participants provided written informed consent after receiving a full explanation of the study procedures.

MBCT program

Our 8‐week MBCT program, based on MBCT for depression by Segal et al., ³⁰ included mindfulness meditation techniques, meta‐cognition training, psychoeducation about panic disorder and its characteristic cognitive distortions, and homework assignments. Treatment sessions were conducted by trained psychiatrists under the direct supervision of a psychiatrist‐in‐chief with extensive MBCT experience. All therapists were educated using a structured manual before delivering the program. Therapists' competency and adherence to the MBCT manual were established by the supervisor. All sessions were monitored for feedback by the supervisor and two independent experienced psychiatrists to ensure treatment integrity. Each treatment group consisted of 6–10 participants. Adherence to MBCT was assessed using a daily homework checklist prior to the start of each session. During homework review, intensive discussions were held with participants to ensure they understood the core contents of the program. The MBCT protocol is described in detail elsewhere. ¹²

MRI acquisition

Diffusion tensor imaging (DTI) data were acquired using a 3.0‐T GE Signa HDxt scanner (GE Healthcare, Milwaukee, WI, USA). No hardware or software changes were made to the scanner during the study period. Three‐dimensional diffusion‐weighted images were acquired using an echo‐planar imaging sequence with the following parameters: (1) repetition time = 17,000 ms; (2) echo time = 108 ms; (3) field‐of‐view = 240 × 240 mm²; (4) matrix = 144 × 144; (5) voxel size = 1.67 × 1.67 × 1.7 mm³; and (5) approximate scan time of 17 min. A double‐echo option was applied to minimize eddy current‐related distortions. To reduce the impact of spatial distortions, an eight‐channel coil was used to perform parallel imaging using ASSET (Array Spatial Sensitivity Encoding Techniques, GE) with a sensitivity‐encoding factor of two. In total, 70 axial slices were acquired for the whole brain in 51 directions with diffusion weighting of b = 900 s/mm², and eight baseline scans were obtained with b = 0 s/mm². Diffusion tensor images were estimated from the diffusion‐weighted images using the least‐squares method.

DTI processing

Raw diffusion images underwent standard preprocessing steps including skull stripping and eddy current correction using the Functional MRI of the Brain (FMRIB) Software Library (FSL version 5.0; Oxford, UK; https://fsl.fmrib.ox.ac.uk/fsl/). FA images were created by fitting a tensor model to the corrected diffusion data and aligned in the Montreal Neurologic Institute (MNI) standard space using the FMRIB's Nonlinear Image Registration Tool (FNIRT). Transformed FA images were combined and applied to the original FA map. A mean FA image was created by averaging all transformed FA images and skeletonized to represent the centers of the white matter tracts. The skeleton was thresholded using an FA >0.2 to include only major fiber bundles.

White matter regions underlying the following cortical midline structures were selected as regions of interest (ROIs) from the left and right hemispheres: (1) anterior cingulate cortex (ACC); (2) dorsomedial prefrontal cortex (dmPFC); (3) ventromedial prefrontal cortex (vmPFC); (4) posterior cingulate cortex (PCC); (5) precuneus. The borders of ROIs were manually drawn on the mean FA skeleton images according to the cortical parcellations of the Automated Anatomical Labeling (AAL) atlas ³¹ and Harvard‐Oxford (HO) atlas. ³² The mean FA values of each ROI were extracted from the individual FA map of study participants for further statistical analysis.

Statistical analysis

Sociodemographic, clinical, and symptom characteristics were compared using analysis of variance and independent t‐test for continuous variables, and chi‐squared test for categorical variables. Linear mixed model analysis was performed for repeated measures. We tested the main effects of group and time and their interaction with fixed effects for the severity of state and trait symptoms. Longitudinal FA changes in each ROI were compared between groups using percentage changes calculated by the formula: 100 × (FA_follow‐up−FA_baseline)/FA_baseline; the main effects of group and hemisphere (i.e., left and right) and their interaction were tested. The significance of longitudinal changes in FA was examined using Wilcoxon signed‐rank test in each group. Using Spearman's rank correlations, we assessed the relationships between percentage changes in FA and difference (Δ) in symptom measure scores, calculated by subtracting symptom scores at the baseline from those at the 2‐year follow‐up. All statistical tests were conducted using SPSS (version 28; IBM Corp., Armonk, NY, USA) with the statistical significance set at P < 0.05. Bonferroni correction was applied for post‐hoc comparisons.

Results

Sociodemographic and clinical characteristics

Table 1 presents the sociodemographic and clinical characteristics of all participants. No significant differences were observed among the three groups in age, sex, education years, and intracranial volume at the baseline. All patients with panic disorder were drug‐naïve at the baseline scan. No significant differences were observed in the duration of pharmacotherapy or daily dose between the patient groups.

Table 1.

Sociodemographic and clinical characteristics of study participants (n = 71)

				Statistical analysis
	HCs (n = 25)	MBCT+PT (n = 26)	PT‐alone (n = 20)	Test	P
Age at baseline (years), mean ± SD	34.4 ± 8.3	37.0 ± 9.5	35.5 ± 11.0	F = 0.46	0.633
Women, n (%)	13 (52.0)	13 (50.0)	12 (60.0)	χ 2 = 0.49	0.783
Education (years), mean ± SD	15.3 ± 1.5	14.7 ± 2.6	14.4 ± 2.0	F = 1.14	0.327
Intracranial volume at baseline (ml), mean ± SD	1546.81 ± 135.50	1519.53 ± 119.77	1489.43 ± 130.92	F = 1.11	0.336
Time between baseline and follow‐up MRI scans (months), mean ± SD	24.4 ± 2.1	25.1 ± 1.3	25.4 ± 1.3	F = 1.90	0.157
Duration of PT between scans (months), mean ± SD		23.7 ± 4.8	22.6 ± 7.0	t = 0.64	0.527
Daily dose of maintenance PT
Antidepressants (mg/day), mean ± SD †		20.6 ± 12.7	23.1 ± 14.4	t = −0.63	0.533
Benzodiazepines (mg/day), mean ± SD ‡		1.1 ± 1.3	1.3 ± 1.4	t = −0.58	0.568

Abbreviations: HCs, healthy controls; MBCT, mindfulness‐based cognitive therapy; PT, pharmacotherapy; SD, standard deviation.

^† Daily antidepressant doses were converted into fluoxetine‐equivalent doses according to Hayasaka et al. (2015). ³³

^‡ Lorazepam‐equivalent doses were calculated for benzodiazepines. Both regular and supplemental administrations were included.

State and trait symptom characteristics

Table 2 and Fig. 1 show the state and trait symptom characteristics and results of the linear mixed model analysis. Significant group‐by‐time interactions were found in the PDSS, BDI‐II, BAI, and fear of respiratory symptoms of the ASI‐R. Compared to HCs, the patient groups exhibited more severe state symptoms of panic, depression, and anxiety at the baseline and follow‐up. No significant differences were observed between the two patient groups at the baseline. After 2 years, when all state symptoms improved in both patient groups, the MBCT+PT group showed significantly lower PDSS scores than the PT‐alone group did. Similar tendencies were observed for the severity of trait symptoms; the two patient groups showed higher symptom scores than HCs. Significant improvements in several trait symptoms were noted in patients after 2 years, including neuroticism and fear of respiratory symptoms, publicly observable anxiety reactions, and cardiovascular symptoms in the MBCT+PT group and fear of respiratory symptoms and cardiovascular symptoms in the PT‐alone group. No significant differences in the severity of trait symptoms were found between the MBCT+PT and PT‐alone groups at the baseline and 2‐year follow‐up.

Table 2.

Longitudinal changes in state and trait symptoms between the baseline and 2‐year follow‐up

	HCs (n = 25)			MBCT+PT (n = 26)			PT‐alone (n = 20)			Statistical analysis
	Baseline	Follow‐up	Δ	Baseline	Follow‐up	Δ	Baseline	Follow‐up	Δ	Group	Time	Group‐by‐time
State symptoms, mean ± SD
PDSS	0.0 ± 0.0	0.1 ± 0.4	0.12 ± 0.44	12.6 ± 5.6	3.0 ± 3.4	−9.54 ± 5.53	12.3 ± 5.7	6.0 ± 5.0	−6.25 ± 5.08	F = 50.97, P < 0.001	F = 103.23, P < 0.001	F = 32.92, P < 0.001
BDI‐II	2.8 ± 2.4	2.9 ± 5.3	0.43 ± 4.60	15.6 ± 7.9	10.1 ± 12.0	−5.00 ± 9.45	18.5 ± 7.8	13.6 ± 9.1	−4.95 ± 8.59	F = 22.40, P < 0.001	F = 12.00, P < 0.001	F = 3.63, P = 0.032
BAI	1.4 ± 1.5	1.7 ± 2.7	0.35 ± 2.60	23.1 ± 10.8	10.6 ± 10.4	−12.72 ± 13.00	25.5 ± 14.1	14.9 ± 11.3	−10.60 ± 13.06	F = 37.11, P < 0.001	F = 35.31, P < 0.001	F = 10.05, P < 0.001
Trait symptoms, mean ± SD
Neuroticism	4.1 ± 2.6	3.6 ± 2.9	−0.57 ± 2.43	8.0 ± 2.9	6.8 ± 3.5	−1.26 ± 2.70	8.1 ± 3.6	7.2 ± 3.5	−0.90 ± 3.26	F = 12.78, P < 0.001	F = 6.43, P = 0.014	F = 0.51, P = 0.604
ASI‐R
Fear of respiratory symptoms	0.2 ± 0.5	0.5 ± 1.4	0.35 ± 1.50	16.8 ± 9.9	8.4 ± 9.6	−8.96 ± 11.24	18.0 ± 9.7	11.1 ± 7.9	−6.85 ± 7.36	F = 32.71, P < 0.001	F = 28.37, P < 0.001	F = 8.77, P < 0.001
Fear of publicly observable anxiety reactions	1.5 ± 3.1	1.5 ± 3.0	0.00 ± 3.26	9.2 ± 6.4	6.7 ± 8.3	−2.57 ± 8.05	10.4 ± 6.6	6.7 ± 6.6	−3.70 ± 5.79	F = 13.70, P < 0.001	F = 8.06, P = 0.006	F = 2.21, P = 0.117
Fear of cardiovascular symptoms	0.4 ± 0.8	0.7 ± 2.1	0.22 ± 2.35	13.0 ± 7.5	7.6 ± 10.5	−5.09 ± 9.79	15.3 ± 11.3	10.9 ± 10.2	−4.40 ± 10.84	F = 21.75, P < 0.001	F = 9.40, P = 0.003	F = 2.90, P = 0.062
Fear of cognitive dyscontrol	0.1 ± 0.6	0.2 ± 0.5	0.04 ± 0.64	4.3 ± 4.6	3.0 ± 4.9	−1.30 ± 4.37	6.0 ± 6.9	4.9 ± 5.8	−1.05 ± 4.39	F = 9.96, P < 0.001	F = 3.15, P = 0.081	F = 0.95, P = 0.390

Abbreviations: ASI‐R, Anxiety Sensitivity Index‐Revised; BDI‐II, Beck Depression Inventory‐II; BAI, Beck Anxiety Inventory; HCs, healthy controls; MBCT, mindfulness‐based cognitive therapy; PT, pharmacotherapy; PDSS, Panic Disorder Severity Scale; SD, standard deviation.

Fig. 1.

Longitudinal changes in the severity of state and trait symptoms between the baseline and 2‐year follow‐up. *Significant differences from the baseline. ^#Significant between‐group differences with HCs. ^##Significant differences between the MBCT+PT and PT‐alone groups. Post‐hoc comparisons were considered significant at P < 0.05 after Bonferroni correction. BDI‐II, Beck Depression Inventory‐II; BAI, Beck Anxiety Inventory; HCs, healthy controls; MBCT, mindfulness‐based cognitive therapy; PT, pharmacotherapy; PDSS, Panic Disorder Severity Scale; SE, standard error.

Treatment outcome after 2 years of follow‐up in patients with panic disorder

Figure 2 shows the serial follow‐up of the severity of panic symptoms using the PDSS in the MBCT+PT and PT‐alone groups. There was a significant main effect of group (F = 7.08, P = 0.011) and time (F = 34.07, P < 0.001) and a significant group‐by‐time interaction effect (F = 5.45, P < 0.001). In detail, the PDSS scores were significantly lower in the MBCT+PT group than in the PT‐alone group at 8 weeks, 6 months, 1 year, and 2 years from the baseline. The MBCT+PT group achieved significant improvement in panic symptoms after 8 weeks of pharmacotherapy with adjuvant MBCT and remained stable for 2 years. On the other hand, the PT‐alone group showed a significant decrease in PDSS scores after 6 months of pharmacotherapy from the baseline. Seventeen out of 26 MBCT+PT patients (65.4%) and 6 out of 20 PT‐alone patients (30.0%) achieved remission according to the Ballenger's criteria. The remission rate after 2 years was significantly lower in the MBCT+PT group than in the PT‐alone group (χ ² = 5.66, P = 0.017).

Fig. 2.

Longitudinal changes in PDSS scores across 2 years of follow‐up. The MBCT group showed early improvement in panic symptoms after 8 weeks of pharmacotherapy with adjuvant MBCT and remained stable for 2 years. In the PT‐alone group, noticeable changes from the baseline in the severity of panic symptoms were observed after at least 6 months of pharmacotherapy. A significant group‐by‐time interaction was found, indicating that the MBCT+PT group showed better outcomes across 2 years of follow‐up than the PT‐alone group. *Significant differences from the baseline. **Significant differences at 2‐year follow‐up. ^#Significant differences between the MBCT+PT and PT‐alone groups. Post‐hoc comparisons were considered significant at P < 0.05 after Bonferroni correction. MBCT, mindfulness‐based cognitive therapy; PDSS, Panic Disorder Severity Scale; PT, pharmacotherapy; SE, standard error.

[Correction added on Apr 12,2023, after first online publication: In the previous version, the prior two sentences were “Fourteen out of 21 MBCT+PT patients (66.7%) and 6 out of 18 PT‐alone patients (33.3%) achieved remission according to the Ballenger's criteria. The remission rate after 2 years was significantly lower in the MBCT+PT group than in the PT‐alone group (χ ² = 4.31, P = 0.038).”]

FA of white matter tracts underlying the CMS at baseline and follow‐up

Table 3 presents the baseline and follow‐up FA of the ROIs and their percentage changes. Cross‐sectional comparisons at each time point were conducted using linear mixed model analysis to test the main effects of group and hemisphere and their interaction. At baseline, there was a significant main effect of hemisphere on the FA of white matter tracts underlying the CMS (ACC: F = 4403.44, P < 0.001; dmPFC: F = 77.82, P < 0.001; vmPFC: F = 28.38, P < 0.001; PCC: F = 258.87, P < 0.001; precuneus: F = 59.28, P < 0.001); however, no significant between‐group differences (ACC: F = 2.08, P = 0.133; dmPFC: F = 1.39, P = 0.255; vmPFC: F = 0.73, P = 0.485; PCC: F = 1.08, P = 0.347; precuneus: F = 1.17, P = 0.316) were observed. For all groups, FA was lower in the left precuneus than in the right precuneus; however, FA was higher in the left hemispheric ROIs than in their right hemispheric counterparts. The dmPFC showed a significant group‐by‐hemisphere interaction (F = 4.82, P = 0.011), whereas the other ROIs did not (ACC: F = 1.37, P = 0.262; vmPFC: F = 1.52, P = 0.225; PCC: F = 0.81, P = 0.450; precuneus: F = 1.46, P = 0.240). The MBCT+PT group had lower FA in the right dmPFC than HCs (uncorrected P = 0.042). However, this significance was lost after Bonferroni correction.

Table 3.

Longitudinal FA changes in white matter regions underlying the CMS in the selected ROIs between the baseline and 2‐year follow‐up

	HCs (n = 25)			MBCT+PT (n = 26)			PT‐alone (n = 20)			Statistical analysis for % change
	Baseline	Follow‐up	% change	Baseline	Follow‐up	% change	Baseline	Follow‐up	% change	Group	Hemisphere	Group‐by‐hemisphere
Anterior cingulate cortex, mean ± SD
Left	0.657 ± 0.029	0.647 ± 0.023	−1.45 ± 3.06	0.638 ± 0.038	0.626 ± 0.035	−1.84 ± 2.81	0.638 ± 0.025	0.643 ± 0.028	0.77 ± 2.62	F = 7.29, P = 0.001	F = 7.26, P = 0.009	F = 0.27, P = 0.766
Right	0.457 ± 0.028	0.453 ± 0.027	−0.77 ± 2.43	0.440 ± 0.044	0.437 ± 0.042	−0.52 ± 2.96	0.450 ± 0.033	0.458 ± 0.035	1.87 ± 3.65
Dorsomedial prefrontal cortex, mean ± SD
Left	0.384 ± 0.036	0.387 ± 0.038	0.81 ± 4.63	0.371 ± 0.038	0.371 ± 0.036	−0.01 ± 3.97	0.371 ± 0.038	0.386 ± 0.039	4.01 ± 4.45	F = 5.06, P = 0.009	F = 1.03, P = 0.314	F = 0.28, P = 0.755
Right	0.371 ± 0.031	0.374 ± 0.031	0.86 ± 4.62	0.349 ± 0.044	0.352 ± 0.043	0.82 ± 5.46	0.362 ± 0.036	0.377 ± 0.039	4.48 ± 5.12
Ventromedial prefrontal cortex, mean ± SD
Left	0.422 ± 0.029	0.425 ± 0.036	0.63 ± 4.49	0.415 ± 0.036	0.410 ± 0.039	−1.00 ± 5.29	0.415 ± 0.040	0.425 ± 0.034	2.88 ± 6.54	F = 1.80, P = 0.173	F = 0.19, P = 0.665	F = 1.95, P = 0.150
Right	0.417 ± 0.030	0.404 ± 0.038	0.26 ± 5.04	0.403 ± 0.041	0.404 ± 0.038	0.66 ± 5.75	0.406 ± 0.039	0.414 ± 0.036	2.28 ± 5.96
Posterior cingulate cortex, mean ± SD
Left	0.511 ± 0.029	0.506 ± 0.025	−0.86 ± 3.15	0.501 ± 0.029	0.500 ± 0.029	−0.24 ± 3.39	0.505 ± 0.028	0.514 ± 0.025	1.86 ± 2.29	F = 5.35, P = 0.007	F = 1.39, P = 0.243	F = 0.44, P = 0.643
Right	0.486 ± 0.028	0.482 ± 0.029	−0.73 ± 3.59	0.472 ± 0.034	0.471 ± 0.032	−0.13 ± 3.53	0.477 ± 0.027	0.489 ± 0.027	2.50 ± 3.67
Precuneus, mean ± SD
Left	0.397 ± 0.042	0.397 ± 0.040	0.04 ± 4.62	0.384 ± 0.044	0.384 ± 0.041	0.36 ± 6.58	0.382 ± 0.045	0.397 ± 0.046	4.18 ± 4.96	F = 3.83, P = 0.027	F = 0.09, P = 0.760	F = 0.31, P = 0.736
Right	0.418 ± 0.041	0.416 ± 0.038	−0.40 ± 4.93	0.396 ± 0.050	0.397 ± 0.046	0.67 ± 6.30	0.402 ± 0.054	0.416 ± 0.051	3.92 ± 6.18

Abbreviations: CMS, cortical midline structures; FA, fractional anisotropy; HCs, healthy controls; MBCT, mindfulness‐based cognitive therapy; PT, pharmacotherapy; ROI, region of interest; SD, standard deviation.

The results at the 2‐year follow‐up were in line with those at baseline except for the ACC; there was a significant main effect of both group (F = 3.15, P = 0.049) and hemisphere (F = 3744.19, P < 0.001) but no significant group‐by‐hemisphere interaction (F = 0.61, P = 0.544). Post‐hoc analysis revealed that the MBCT+PT group had lower FA than HCs in the left ACC (uncorrected P = 0.028) and lower FA than the PT‐alone group in the right ACC (uncorrected P = 0.034), although the significance was lost after Bonferroni correction. The other ROIs showed a significant main effect of hemisphere (dmPFC: F = 56.82, P < 0.001; vmPFC: F = 27.32, P < 0.001; PCC: F = 246.91, P < 0.001; precuneus: F = 71.45, P < 0.001). However, there were no significant between‐group differences (dmPFC: F = 2.31, P = 0.107; vmPFC: F = 1.18, P = 0.314; PCC: F = 2.01, P = 0.141; precuneus: F = 1.13, P = 0.329) and group‐by‐hemisphere interactions (dmPFC: F = 2.96, P = 0.059; vmPFC: F = 0.96, P = 0.388; PCC: F = 1.14, P = 0.325; precuneus: F = 0.89, P = 0.416). The statistical significance of the results remained unchanged when the duration of pharmacotherapy and daily maintenance doses between the two scans were included as covariates at the 2‐year follow‐up.

Longitudinal FA changes in white matter tracts underlying the CMS 2 years following MBCT

The percentage change in FA was significantly different between groups in the ACC, dmPFC, PCC, and precuneus. Inter‐hemispheric differences in the FA percentage change were only significant in the ACC. No group‐by‐hemisphere interactions were found in any of the ROIs (Table 3). As demonstrated in Fig. 3, the significant main effects of group were driven by the higher percentage changes in FA in the PT‐alone group than in the MBCT+PT group and HCs. The FA percentage changes did not correlate with the maintenance dose of antidepressants and benzodiazepines and the duration of pharmacotherapy in patients with panic disorder (r < 0.252, P > 0.122), and the results were largely similar after controlling for those variables. The results of Wilcoxon signed‐rank test indicated that the percentage changes in FA for 2 years were driven by a decreased FA in the left ACC in the MBCT+PT group and increased FA in the bilateral dmPFC, PCC, and precuneus in the PT‐alone group.

Fig. 3.

White matter regions of interest underlying the cortical midline structures and their percentage changes in FA at 2 years. *Significant 2‐year differences at P < 0.05; **P < 0.01; ***P < 0.001 (Wilcoxon signed‐rank test). ^#Significant between‐group differences at P < 0.05; ^## P < 0.01 (linear mixed model analysis; Bonferroni‐corrected for multiple comparisons). ACC, anterior cingulate cortex; dm/vmPFC, dorsomedial/ventromedial prefrontal cortex; FA, fractional anisotropy; HCs, healthy controls; MBCT, mindfulness‐based cognitive therapy; PCC, posterior cingulate cortex; PT, pharmacotherapy; SE, standard error.

Correlations between white matter FA of the CMS and symptoms

Exploratory correlation analysis was performed for each group of patients between the Δ scores of state and trait symptoms and percentage changes in FA of the left and right CMS ROIs with significant between‐group differences (Fig. 4). In the MBCT+PT group, the improvement of state and trait symptoms were associated with FA reduction in the white matter tracts underlying the CMS. The PDSS Δ score was positively correlated with FA percentage changes in the left ACC (ρ = 0.456, P = 0.019) and right PCC (ρ = 0.398, P = 0.044). The Δ score of fear of respiratory symptoms showed positive correlations with FA percentage changes in the left ACC (ρ = 0.426, P = 0.030), right ACC (ρ = 0.492, P = 0.011), left PCC (ρ = 0.658, P < 0.001), right PCC (ρ = 0.583, P = 0.002), left precuneus (ρ = 0.410, P = 0.038), and right precuneus (ρ = 0.463, P = 0.017). The Δ score of fear of publicly observable anxiety reactions was associated with FA percentage changes in the right ACC (ρ = 0.455, P = 0.020). On the other hand, distinct patterns of association were observed in the PT‐alone group compared to the MBCT+PT group. The PT‐alone group showed no significant associations between the Δ score of the PDSS and FA of the CMS ROIs. Instead, positive correlations were found between the Δ score of neuroticism and FA percentage changes in the left precuneus (ρ = 0.472, P = 0.036) and between the Δ score of fear of publicly observable anxiety reactions and FA percentage changes in the left dmPFC (ρ = 0.449, P = 0.047).

Fig. 4.

Associations between percentage changes in FA and the difference scores of state and trait symptoms in each group of patients with panic disorder. Numbers in each cell are Spearman's ρ, and the black boxes indicate statistically significant associations (uncorrected P < 0.05). ACC, anterior cingulate cortex; BDI‐II, Beck Depression Inventory‐II; BAI, Beck Anxiety Inventory; dmPFC, dorsomedial prefrontal cortex; FA, fractional anisotropy; MBCT, mindfulness‐based cognitive therapy; PT, pharmacotherapy; PDSS, Panic Disorder Severity Scale; PCC, posterior cingulate cortex.

Discussion

To our knowledge, this is the first prospective study to demonstrate the long‐term benefits of MBCT on white matter tracts underlying the CMS in patients with panic disorder undergoing PT. Adjuvant MBCT with pharmacotherapy resulted in neuroplastic changes in white matter as well as improved clinical outcomes over 2 years. Significant differences in FA changes were found at the 2‐year follow‐up among the HC, MBCT+PT, and PT‐alone groups, suggesting that each group followed distinctive trajectories in white matter neuroplasticity. The HCs showed no significant changes, whereas the MBCT+PT group exhibited decreased FA in the left ACC, and the PT‐alone group showed increased FA in the bilateral dmPFC, PCC, and precuneus. Taken together with the positive association between the percentage of FA reduction and decreased severity of state and trait symptoms, our findings highlight the benefits of MBCT in remodeling pathological white matter connections, leading to improvements in clinical symptoms and trait vulnerability in patients with panic disorder.

In this study, patients who received MBCT adjuvant to pharmacotherapy showed rapid symptomatic improvements after 8 weeks of sessions, and this effect lasted for 2 years. Although the PT‐alone group also showed improvements in panic symptoms, the first significant decrease in PDSS scores from baseline was noted after 6 months of pharmacotherapy. Furthermore, the MBCT+PT group had less severe panic symptoms and fewer relapses throughout the entire follow‐up period than the PT‐alone group did. This aligns with the results of our previous studies with independent samples that showed the short‐ and long‐term effectiveness of adjuvant MBCT for patients with panic disorder. ¹⁰ , ¹¹ , ¹² Although antidepressants are helpful in reducing the intensity and frequency of panic attacks, many patients still exert substantial effort into cognitive restructuring to overcome panic‐related catastrophic apprehension, such as generalized worry, fear of physical illness, and agoraphobia. ³⁴ In this case, mindfulness provides an alternative way of relating with thoughts and feelings by training the self‐regulation of attention to focus on the here and now, rather than directly countering maladaptive cognition and behaviors. ⁹ MBCT for panic disorder enables patients to adopt a non‐judgmental and accepting attitude to the reality of the present moment as it is and act on automatic pilot with mindful awareness. Therefore, MBCT could be an effective strategy to optimize short‐ and long‐term treatment outcomes of patients with panic disorder, particularly when combined with pharmacotherapy.

After 2 years, HCs, MBCT+PT patients, and PT‐alone patients showed different patterns of white matter FA changes in the CMS. In HCs, there were no significant FA differences between the baseline and 2‐year follow‐up. Given that age‐related decline in white matter volume and integrity is accelerated around the fifth decade of life, ³⁵ , ³⁶ , ³⁷ the 2‐year‐long observation of white matter connectivity may not have been sufficient to detect changes in our HCs who were in their mid‐30s. On the other hand, the MBCT+PT and PT‐alone groups showed significant longitudinal FA changes in opposite directions, although panic symptoms and concomitant anxiety and depression improved after 2 years in both groups. FA was decreased in the left ACC in the MBCT+PT group and increased in the bilateral dmPFC, PCC, and precuneus in the PT‐alone group. Although these changes did not result in significant differences in cross‐sectional comparison among the three groups at the 2‐year follow‐up, considerable longitudinal changes in white matter connectivity were indicated by significant between‐group differences in FA percentage changes in those regions. Since the patients underwent maintenance pharmacotherapy between the two MRI scans, the longitudinal changes in white matter may be related to medication. Several studies reported that serotonergic antidepressants altered FA or diffusivity of white matter tracts in clinical populations. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ However, the effect of antidepressants on white matter connectivity remains controversial because negative findings also exist. ⁴² , ⁴³ There were no significant correlations between FA percentage changes and the dose or duration medication in our patient groups. Therefore, considering that all patients in the MBCT+PT and PT‐alone groups received pharmacotherapy under similar clinical conditions, the difference in FA changes could be attributable to the adjuvant administration of MBCT.

Neuroimaging studies have provided biological evidence on the ability of mindfulness to promote neuroplasticity, especially in the brain regions processing internal feelings and thoughts such as the cingulate, insular, and medial prefrontal cortices. ¹⁴ , ¹⁵ , ⁴⁴ , ⁴⁵ , ⁴⁶ DTI studies reported enhanced white matter integrity in areas adjacent to the ACC after short‐term medication in healthy populations. ¹⁸ , ¹⁹ At a glance, this seems contradictory to our findings of decreased FA in the ACC and non‐increased FA in the dmPFC, PCC, and precuneus in the MBCT+PT group. The discrepancy with previous results can be attributed to the inherent difference between clinical and non‐clinical populations. High or increased FA is commonly considered to reflect well‐integrated, healthy white matter structure, and low FA has been observed in white matter tracts of the fronto‐temporo‐limbic circuitry in patients with panic disorder. ⁴⁷ , ⁴⁸ , ⁴⁹ However, as with many biological phenomena observed in nature, a high value does not necessarily indicate improvement. Decreased white matter FA in the ACC is associated with less severe panic symptoms, ⁵⁰ and patients with panic disorder who respond to antidepressant therapy exhibit lower fronto‐temporo‐limbic FA than treatment non‐responders. ⁵¹ , ⁵² The positive effect of decreased FA is also supported by our findings where the percentage decrease in FA was associated with the improvement of state and trait symptoms in patients with panic disorder. This relationship was more pronounced in the MBCT+PT group; patients in the MBCT+PT group who had greater FA decrease in the left ACC showed better improvement in the severity of panic symptoms and anxiety sensitivity to respiratory symptoms than the PT‐alone group. Although FA in other regions did not change significantly over time, FA in the PCC and precuneus also showed similar relationships. Patients in the PT‐alone group, who had increased FA in the dmPFC, PCC, and precuneus, showed no significant associations between FA and symptom reductions, except between the FA in the left dmPFC and anxiety sensitivity to publicly observable anxiety reactions and between the FA in the left precuneus and neuroticism.

The ACC is involved in the internal monitoring and adaptive control of behavior in conjunction with cognitive, affective, and visceromotor functions. ⁵³ It is also implicated in producing warning signals such as pain and anxiety to avoid future harm in response to errors and threats. ⁵⁴ The dmPFC, PCC, and precuneus are major components of the default mode network, which is active when attention is allocated to intrinsic feelings and thoughts. ⁵⁵ Each region is suggested to play a distinct role in self‐related mental operations; the dmPFC is involved in the evaluation and judgment of self‐referential stimuli, whereas the PCC and precuneus integrate self‐referential stimuli in the context of one's own person. ²⁰ Functional and structural connectivity in these regions is increased in patients with depressive and anxiety disorders who are preoccupied with their inner selves. ⁵⁶ , ⁵⁷ , ⁵⁸ Furthermore, pathological anxiety is associated with hyperactivity of brain regions in the salience and default mode networks, suggesting the increased brain burden of coping with fearful inner experiences and unwarranted worry. ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² Excessive neuronal excitation leads to the maladaptive formation of redundant axonal components or other glial cells, leading to aberrant communication between brain regions. ⁶³ , ⁶⁴ , ⁶⁵ In this regard, it could be speculated that increased FA in the dmPFC, PFC, and precuneus in PT‐alone patients reflects the cognitive burden of directly countering maladaptive cognition and behaviors related to panic symptoms, even though their symptoms had improved after pharmacotherapy. On the other hand, mindfulness fundamentally alters the primary relationship to anxiogenic experiences in a decentering and non‐judgmental manner and reduces hypervigilance toward inner experiences. Functional MRI studies have shown that mindfulness does not simply increase or decrease brain activity, but rather modulates brain connectivity of the corticolimbic circuits, resulting in well‐organized functioning in clinical samples. ⁶⁶ Therefore, mindfulness, which helps patients reestablish the ability to navigate the world without fear and anxiety, could play a pivotal role in loosening the overtightened brain and reducing the cognitive distress of handling excessive anxiety and erroneous beliefs related to panic symptoms. ⁶⁷ , ⁶⁸ , ⁶⁹ Although further research is required, the attenuation of excessive white matter connectivity in the CMS after MBCT can be considered a possible mechanism underscoring the effect of mindfulness on panic symptoms and underlying traits.

Some limitations should be considered when interpreting our results. First, the small sample size is a major limitation that prevented us from detecting statistically significant changes and expanding the investigation to the whole brain. Although future studies with larger sample sizes are necessary, our study still makes a meaningful contribution to the literature, as it is the first to investigate the long‐term benefits of MBCT in patients with panic disorder who were on maintenance pharmacotherapy. Second, this study did not randomize patients into groups with and without MBCT. To minimize selection bias resulting from researchers, we ensured that therapists who delivered MBCT were blinded to the purpose and hypotheses of the study and patients' enrollment. In addition, researchers who contacted the patients every month were blinded to the group assignment. However, it might not be enough to rule out selection bias from the patients' side. Therefore, double‐blinded, randomized controlled studies are needed to confirm our findings. Third, the current study focuses on patients with panic disorder who required maintenance pharmacotherapy, thus weakening the generalizability of our results to patients with less severe panic disorder. The use of antidepressant and anxiolytic medications during maintenance therapy may have exerted confounding effects on white matter connectivity; however, our patients with panic disorder received pharmacotherapy according to standardized protocols, and no significant differences were observed in the duration of pharmacotherapy and daily maintenance doses between the two patient groups. Therefore, our findings support the additive effects of MBCT on long‐term neuroplastic changes in the brains of patients with panic disorder. In addition, considering the broad applicability of mindfulness for populations with anxiety and depression, the long‐term benefits of MBCT are likely applicable to patients with subthreshold or mild panic disorder.

In conclusion, this study demonstrated the long‐term benefits of MBCT as an adjunct to pharmacotherapy in normalizing excessive connectivity of white matter regions underlying the CMS in patients with panic disorder. At the symptom level, maintenance pharmacotherapy alone was effective in improving panic symptoms; however, patients with panic disorder who received MBCT exhibited better clinical outcomes and positive neuroplastic changes that were associated with improved panic symptoms and trait vulnerability after 2 years. Our findings provide clinical and neurobiological evidence for the clinical applicability of MBCT as a robust treatment for patients with panic disorder.

Disclosure statement

The authors declare that there is no conflict of interest in relation to this study.

Author contributions

S.H.L. designed the study. B.K., K.S.L., T.K.C., and S.H.L. managed the participant recruitment and data acquisition. M.B., B.K., and S.H.L. compiled the database and conducted the data preprocessing and statistical analyses. M.B. and S.H.L. implemented the literature reviews and interpretation of data. M.B. wrote the first draft of the manuscript, and S.H.L. provided the critical revision of the manuscript. All authors contributed to and approved the final manuscript.