Effect of Low-Intensity Transcranial Focused Ultrasound Stimulation in Patients With Major Depressive Disorder: A Randomized, Double-Blind, Sham-Controlled Clinical Trial
Article information
Abstract
Objective
Low-intensity transcranial focused ultrasound (tFUS) has emerged as a promising non-invasive brain stimulation modality with high spatial selectivity and the ability to reach deep brain areas. The present study aimed to investigate the safety and effectiveness of low-intensity tFUS in treating major depressive disorder.
Methods
Participants were recruited in an outpatient clinic and randomly assigned to either the verum tFUS or sham stimulation group. The intervention group received six sessions of tFUS stimulation to the left dorsolateral prefrontal cortex over two weeks. Neuropsychological assessments were conducted before and after the sessions. Resting-state functional magnetic resonance imaging (rsfMRI) was also performed to evaluate changes in functional connectivity (FC). The primary outcome measure was the change in depressive symptoms, assessed with the Montgomery–Åsberg Depression Rating Scale (MADRS).
Results
The tFUS stimulation sessions were well tolerated without any undesirable side effects. The analysis revealed a significant main effect of session sequence on the MADRS scores and significant interactions between the session sequences and groups. The rsfMRI analysis showed a higher FC correlation between the right superior part of the subgenual anterior cingulate cortex (sgACC) and several other brain regions in the verum group compared with the sham group.
Conclusion
Our results reveal that tFUS stimulation clinically improved MADRS scores with network-level modulation of a sgACC subregion. This randomized, sham-controlled clinical trial, the first study of its kind, demonstrated the safety and probable efficacy of tFUS stimulation for the treatment of depression.
INTRODUCTION
Major depressive disorder (MDD) is a prevailing psychiatric disorder with high recurrence and suicide rates [1]. Antidepressants are considered the first-line treatment for MDD; however, a significant number of patients do not respond to antidepressants or reach full remission [2]. Therefore, non-pharmacological treatments, including brain stimulation techniques such as transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS), have been investigated in treating MDD [3,4]. TMS, which has demonstrated great efficacy over an extended period, and tDCS, while showing some promising effects, require further validation for robust effectiveness in treating MDD. Regarding TMS, however, the spatial resolution of the techniques remains limited although the effects can reach deep brain regions using double-cone or H coils [5,6]. To overcome these limitations, research has sought a new stimulation modality with better spatial specificity and the potential to reach deep brain areas.
Transcranial focused ultrasound (tFUS) has risen as a novel non-invasive brain stimulation technique. Low-intensity tFUS delivers non-thermal, focused acoustic pressure waves to specific subregions of the brain through the skull [7]. The safety and neuromodulatory potentials of tFUS have been actively investigated in small [8] and large animals [9], including non-human primates [10], and stimulation of sensorimotor and visual brain areas among healthy human volunteers has shown excellent safety [11]. Recent studies have shown the clinical utility of modulating regional brain activity with low-intensity tFUS to treat consciousness disorders or suppress epilepsy [12,13]. Thus, tFUS is expected to provide potential treatment options for MDD [14] by stimulating specific brain regions associated with known brain circuit dysfunction in depressed patients.
Previous neuroimaging investigations of patients with MDD have revealed that the risk of occurrence and severity of depression are linked to decreased activity in the left prefrontal cortex [15]. Further neuroimaging studies have identified asymmetric activity in the prefrontal cortex associated with depressive symptoms [16]. Electroconvulsive therapy studies have also associated its therapeutic effect on depression with modulation of activity in the left prefrontal cortex, especially in the dorsolateral prefrontal cortex (DLPFC) [17]. These results were supported by later TMS and tDCS studies on treating depression by stimulating the DLPFC [18,19].
Although TMS and tDCS stimulation of the left DLPFC have demonstrated their effects, it should be noted that stimulation target consistency and focality of stimulation may still be limited at present [20,21]. Weigand et al. [22] conducted functional connectivity (FC) analysis of resting-state functional magnetic resonance imaging (rsfMRI) associated with image-guided TMS stimulation of MDD patients and found that subregions within the left DLPFC yielded mixed therapeutic responses for MDD, which may help explain the subpar antidepressive efficacies. Moreover, the authors identified a specific DLPFC subregion (the left DLPFC, Montreal Neurological Institute [MNI] coordinates: -42, 44, 30) for more effective TMS treatment in addition to strong functional anti-correlation of DLPFC subregions with the subgenual anterior cingulate cortex (sgACC) [22]. The existence of a subregion in the DLPFC that resists selective stimulation with existing non-invasive brain stimulation modalities warrants a new technique with higher spatial resolution. As tFUS can sonicate very small brain areas, we were motivated to use it to stimulate this subregion of the left DLPFC of MDD patients.
We hypothesized that multiple tFUS stimulation sessions would be safe and yield anti-depressive effects among patients with MDD compared with a sham-treated group. To test our hypothesis, neuropsychological assessments, including depressive symptom scales and neurocognitive function tests, were conducted before and after the stimulation sessions. The primary outcome was the change in depressive symptoms as assessed by the Montgomery–Åsberg Depression Rating Scale (MADRS) [23]. We also examined the changes in other clinical symptomatic variables and neurocognitive functions as secondary outcomes. In addition, we performed rsfMRI to evaluate the spatiotemporal features of FC associated with the tFUS sessions.
METHODS
Overall experimental design and patient selection
The present clinical trial (clinicaltrials.gov ID: NCT04405791) was randomized, double-blind, and sham-controlled. The authors assert that 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. All procedures involving human subjects/patients were approved by the Institutional Review Board of Gangnam Severance Hospital in Seoul, Korea (No. 3-2018-0352). The Korean Ministry of Food and Drug Safety also approved this trial (No. 910). The results presented in this study are findings from an exploratory clinical trial conducted prior to a confirmatory clinical trial.
The overall experimental procedure is illustrated in Figure 1A. Participants with MDD were recruited through both public advertisements and postings within the psychiatric outpatient clinics of Gangnam Severance Hospital. Written informed consent was obtained from all recruits who were deemed eligible for the trial based on the inclusion/exclusion criteria listed in Table 1. The patients in this study were not specifically restricted beyond these inclusion/exclusion criteria, therefore, it is not exclusively comprised of treatment-resistant patients. The participants underwent brain magnetic resonance imaging (MRI) and computed tomography (CT) and responded to a detailed survey of their past medical history. CT scans were needed to obtain information on the skull structure including skull thickness, extensive calcification within the sonication pathway.
Study design and experimental procedures. A: Sequence of the experimental procedure. B: Description of the tFUS setup. Illustration of the tFUS setup and headgear on a plastic mannequin head. C: Mapping of acoustic focus perpendicular to and along (inset) the sonication path. The white dashed line indicates the ellipsoidal FWHM profile. The FUS focus at FWHM intensity was 8.5 mm in diameter and 51 mm in length according to the characterization of the acoustic intensity profile (-6 dB pressure, 50% intensity), covering a transverse (20×20 mm2 with 0.5-mm step) and a longitudinal plane (60×100 mm2 with 0.5-mm step) perpendicular to the beam path at the point of maximum intensity. The center of the maximum intensity area was 30 mm away from the exit plane of the transducer on the beam path. The white solid line denotes -3 dB pressure (75% intensity). D: Enrollment diagram and flowchart of this study. Out of 40 initial recruits, seven were excluded (3 not meeting MDD criteria, 2 with MRI contraindications, 1 with a low MADRS score, N=1 with a low MADRS score, and N=1 with sudden alteration of the antidepressant regimen). Three more were excluded for schizophrenia, intellectual disability, and cavernous malformation. Four declined participations before randomization. The remaining 26 participants were assigned to verum (N=13) or sham (N=13) groups. Of these subjects, two discontinued, and the remaining 24 completed all tFUS sessions. One subject who completed the stimulation procedure could not complete whole study procedure due to a recent layoff issue unrelated to this study. Thus, 23 patients (N=11 verum, N=12 sham) finished the study. Four subjects’ rsfMRI data were excluded for quality issues, leaving 19 for resting-state FC analysis. MRI, magnetic resonance imaging; CT, computed tomography; tFUS, transcranial focused ultrasound; Wk, week; fMRI, functional magnetic resonance imaging; FWHM, full-width at halfmaximum; MDD, major depressive disorder; MADRS, Montgomery–Åsberg Depression Rating Scale; rsfMRI, restingstate fMRI; FC, functional connectivity.
Inclusion and exclusion criteria for low-intensity transcranial focused ultrasound stimulation clinical trial for patients with major depressive disorder
Drawing upon previous research [24-26] and aiming for 80% statistical power with a significance level of p<0.05, it was deduced that a minimum of 27 participants per group would be requisite. Accounting for a 10% potential dropout rate, enrolling a total of 60 participants was estimated to ensure robust sample size. Nonetheless, acknowledging the exploratory essence of the trial, the determination was made to enroll 40 participants.
The participants were then randomly assigned to either the verum group or the sham group. The randomization table was generated and implemented using the website (http://www.randomizer.org/) to prevent bias towards one treatment group. To ensure this, initially 10 sets were created, each consisting of 6 participants. This randomization table was pre-constructed with a larger capacity than the actual number of participants. Within each set, an equal 1:1 random allocation was conducted between the verum group and the sham group, and sequential utilization of each set was employed to conduct random allocation. They underwent a battery of baseline neuropsychological assessments (“Baseline,” Figure 1A), which consisted of subtests in the Cambridge Neuropsychological Test Automated Battery (CANTAB) that evaluate changes in depression-related cognitive functions [27]. Depressive symptoms were evaluated using the MADRS and the Quick Inventory of Depressive Symptomatology-Self Report (QIDS-SR) [28]. Anxiety was assessed using the State-Trait Anxiety Inventory (STAI) [29]. The degree of suicidal ideation was measured using the Scale for Suicide Ideation (SSI). Lastly, mood states were evaluated using the Korean edition of the Profile of Mood States (K-POMS) [30]. The MADRS and SSI assessments, which were conducted during eligibility screening, were omitted from baseline neuropsychological assessments. All assessments were conducted by certified psychologists and psychiatrists who were blinded to the group allocation. During the trial period, headaches, heating sensation or sensitivity on the scalp, brain edema, and microhemorrhages were taken into consideration as potential adverse events. These aspects were systematically monitored in participants every two weeks by psychiatrists.
The FUS stimulation sessions (20 min/session) were then performed for each individual thrice a week over two weeks. Two sets of additional neuropsychological assessments were conducted 1 day and 2 weeks after the completion of the FUS stimulation sessions (“End of Treatment” and “Follow-up,” respectively; Figure 1A). A second follow-up MRI was administered on the same day as follow-up neuropsychological assessments. Response rate was determined as the percentage of patients showing at least a 50% reduction in MADRS at “Follow-up” compared to “Baseline,” while remission rate was calculated as the MADRS score of less than nine [31].
Neuroimaging sessions
The brain CT (Revolution HD; GE Healthcare, Chicago, IL, USA) data were obtained with 0.23-mm reconstructed spatial resolution and 0.28-s rotation time. Four adhesive fiducial markers were placed on non-planar locations over the patient’s head (Figure 1B). The fiducial markers and their spatial coordinates in the MRI/CT data were later used for coregistration between the physical and virtual spaces to align the FUS focus to the targeted DLPFC area [32]. For MRI, T1-weighted high-resolution anatomical imaging and rsfMRI information were obtained using a 3-tesla MR scanner (Ingenia MR systems CX; Philips, Amsterdam, The Netherlands). Participants were instructed to rest with their eyes closed during the scan. T1-weighted images were acquired using a spoiled gradient-echo sequence (matrix=224×224, echo time=4.6 ms, repetition time=9.9 ms, field of view=224 mm, slice thickness=1 mm, flip angle=8°). rsfMRI was performed using gradient-echo echo-planar imaging sequences (matrix=80×75, echo time=30 ms, repetition time=2,000 ms, field of view=220 mm, slice thickness=3 mm, flip angle=90°) for 5 min.
Image-guided navigation of FUS transducer and tFUS stimulation
The CT and MRI data were converted to a Digital Imaging and Communications in Medicine file format (using MRIcro 1.40 [https://www.nitrc.org/projects/mricro/] and Amide 1.0.2 [https://amide.sourceforge.net]) and co-registered (using 3D Slicer 4.8.0; https://www.slicer.org/) to provide the individual-specific target area (the left DLPFC, MNI coordinates: -42, 44, 30) [22]. Using statistical parametric mapping software (SPM12 [https://www.fil.ion.ucl.ac.uk/spm]), the individuals’ T1 MRI data were normalized to the MNI space. The anatomical feature corresponding to the targeted coordinate was subsequently identified from each individual’s neuroanatomy.
The image-guided tFUS device was used for the stimulation (NS-US100; Neurosona Co. Ltd, Seoul, Korea, operating at 250-kHz frequency). The spatial profile of the acoustic focus generated by the FUS transducer was characterized within a degassed (<3 ppm) water tank using a needle hydrophone mounted to a three-axis positioning stage. The detailed acoustic field map is shown in Figure 1C.
At least one psychiatrist and an additional research staff member participated in each tFUS session. First, the fiducial markers were placed in the same locations as in the MRI/CT session. Their locations were co-registered to the virtual MRI/CT space. The head motion and its relation to the FUS transducer were tracked in real-time using an infrared camera producing three-dimensional volumetric data (Figure 1B). The sonication path was intended to enter the skull as perpendicularly as possible at the entry point while avoiding the sinus along the beam path. The average distance from the scalp entry to the stimulation target was 16.7±2.2 mm in the verum group and 18.4±2.9 mm in the sham group. A compressible polyvinyl alcohol hydrogel was inserted between the transducer and the scalp to provide acoustic coupling [32]. The hair over the scalp was not shaved but rather carefully combed away from the sonication entry point. Generic ultrasound gel was applied to all interfaces [32]. In the sham group, the same stimulation procedure was repeated without providing actual sonication, and therefore, both groups of patients are kept blind to their respective conditions.
The incident acoustic intensity and pressure at the FUS focus were 3 W/cm2 spatial-peak pulse-average acoustic intensity (Isppa) with a peak negative pressure (Pr) of 300 kPa. The tone burst duration was 1 ms at 50% duty cycle for a duration of 300 ms. We chose this parameter set based on previous studies of healthy human volunteers [11], which yielded robust stimulatory responses. Each sonication was delivered every 6 s for 20 min. With a derating factor of 55% reduction in pressure transmission by the human skull, the estimated in-situ Pr was approximately 135 kPa (corresponding mechanical index of 0.27) with an in-situ acoustic intensity of 600 mW/cm2 spatial-peak temporal average intensity (corresponding spatial-peak temporal average intensity of 300 mW/cm2). Subjects were asked to report any sensations immediately after each session. Following sonication, we retrospectively performed numerical simulation of acoustic wave propagation through the skull to estimate the on-site acoustic intensity and spatial accuracy of the tFUS stimulation [12].
Preprocessing of rsfMRI data
Preprocessing included head motion correction, slice-timing correction, co-registration, normalization, and smoothing with an 8-mm full-width at half-maximum Gaussian filter using the standard pipeline provided in the FC toolbox (CONN; https://web.conn-toolbox.org/) v.18a and SPM12. To mitigate the effects of head movement, images with framewise displacement above 0.9 mm or global BOLD signal changes above 5 standard deviation thresholds were removed using the CONN toolbox (Whitfield-Gabrieli and Nieto-Castanon, 2012 [33]). The signals estimated from white matter and cerebrospinal fluid and realignment parameters were linearly regressed out as confounds in a first-level analysis. The bandpass filter was applied from 0.008 to 0.09 Hz. This denoising step addressed the confounding effects of the participants’ movement without regressing the global signal or affecting intrinsic FC [33]. The data quality of fMRI data was assessed using following parameters: number of invalid scans (with an outlier threshold for scan-to-scan motions at 0.5 mm), maximum motions (mm), mean motions (mm), maximum global signal change (%), mean global signal change (%).
FC analysis
The FC analyses were conducted using a seed-based approach. First, a seed region 5 mm in radius was defined at the sonication target (left DLFPC; -42, 44, 30) [22] to analyze whole-brain connectivity strengths with the stimulation site. Furthermore, according to a previous neuroimaging study on sgACC-based connectivity in patients with depression [34], the seed region in the sgACC was further divided into four subregions of a 3-mm radius in the left superior (sgACCLS), left inferior (sgACCLI), right superior (sgACCRS), and right inferior (sgACCRI) regions [34].
Using the CONN, seed-to-voxel FC maps were generated for each participant. Voxel-wise correlations were calculated between the BOLD signal time series from the seed regions with respect to the whole-brain volume. Correlation coefficients were converted to Z-values using Fisher’s r-to-Z transformation to estimate the FC strengths. We performed t-tests to evaluate the within-group and between-group differences in FC strengths. All imaging analyses were corrected for multiple comparisons using a combination of voxel-level thresholds (p<0.001) and a cluster extent threshold false discovery rates (p<0.05). Pearson’s correlation analyses were conducted between voxel-wise FC strengths (beta values extracted from each participant’s results in the clusters showing a significant within-group or between-group difference) and the changes in MADRS scores during tFUS treatment.
Statistical analysis of demographic and clinical variables
Demographic and clinical data were analyzed using SPSS software, version 18 (SPSS Inc., Chicago, IL, USA). Mann–Whitney tests and chi-squared tests were utilized to investigate differences in baseline demographic and clinical variables between the two groups (verum versus sham) for continuous (age, years of education, Korean Wechsler Adult Intelligence Scale [K-WAIS]) and categorical variables (sex, use of antidepressants), respectively. Mann–Whitney test was selected considering its applicability to analyze data with small sample sizes. Other neuropsychological scores (CANTAB, MADRS, QIDS-SR, STAI, SSI, K-POMS) were analyzed for differences between the verum versus sham treatment using repeated measures analysis of variance (ANOVA). A significance level of p<0.05 was used to determine statistically significant results.
RESULTS
Subject characteristics, stimulation tolerance, and treatment outcome measures
Figure 1D details the study enrollment process and participant characteristics. The enrolled participants (11 verum and 12 sham) tolerated the tFUS sessions well without any undesired effects including headache, heating sensation on scalp or hypersensitivity reaction, cerebral edema, and microhemorrhage. No instances of missing data were observed in both the neuropsychological assessments and fMRI results. Among the 11 subjects (32.4±11.2 years old) in verum group, 5 were male and 6 were female. Five were male and 7 were female among 12 subjects (39.6±12.3 years old) in sham group. Each group has 14.3±1.7 (verum group) and 15.3±3.1 (sham group) years of education, respectively. In the chi-square test and Mann–Whitney test, there was no significant difference in sex distribution, age, years of education, onset age, number of MDD episodes, duration since the onset of MDD and KWAIS, respectively between verum and sham groups. However, a greater proportion of patients were using antidepressants at baseline in verum group (p=0.035). The medications information used by 10 patients in the verum (n=6) and sham group (n=4) are as follows: v1, bupropion 150 mg; v2, sertraline 50 mg, mirtazapine 7.5 mg; v3, venlafaxine 150 mg; v4, agomelatine 50 mg, vortioxetine 10 mg, bupropion 150 mg; v5, venlafaxine 150 mg, fluoxetine 50 mg, bupropion 150 mg; v6, bupropion 150 mg; s1, escitalopram 5 mg, mirtazapine 7.5 mg; s2, vortioxetine 10 mg; s3, duloxetine 60 mg, bupropion 150 mg; and s4, duloxetine 30 mg (Table 2).
Comparison of demographic characteristics and changes in clinical symptoms between the verum and sham treatment groups
The changes in MADRS scores across sessions (primary outcome) are illustrated in Figure 2A, Supplementary Figures 1 and 2 and shown in Table 2 and Supplementary Table 1. The mean MADRS score in the verum group was 28.5±8.4 at baseline; it decreased after stimulation (16.8±6.8 at end of treatment). Further reduction is shown two weeks after the completion of the sessions (14.8±7.2 at follow-up). In the sham group, the initial mean MADRS score (29.2±8.3 at baseline); it also decreased immediately after stimulation (25.7±9.0 at end of treatment) and two weeks after the completion of the sham sessions (24.8±9.3 at follow-up). The main effect of session sequence in both groups were significant (F1,21=23.7, p< 0.001). Although the MADRS scores showed decreasing trends in both the verum and sham stimulation groups, the verum group showed a greater MADRS reduction in time (repeated measures ANOVA across session sequence between the groups; group×session sequence interaction effect F1,21=9.0, p=0.003).
Changes in MADRS scores and results from functional connectivity analysis. A: Changes in MADRS score from the first neuropsychological assessment battery (Baseline) to the third (Follow-up). For each group (verum and sham), the mean changes in MADRS scores from baseline to the end of this study are presented. The blue line indicates the mean change in MADRS scores in the verum group (N=11), and the orange line indicates that in the sham group (N=12), with error bars showing the standard errors. B and C: The brain areas showing higher resting-state FC with respect to the right sgACCRS in the verum group compared with the sham group, displayed on 3D-rendered flattened brain surface (top: right anterior view, bottom: bottom view) (B) and axial anatomical images (C). FC to the sgACCRS showed a significant degree of tFUS-mediated change between the verum and sham condition in 1) the left medial prefrontal cortex (-24, 46, -6), 2) the left middle frontal gyrus (-12, 40, 30), 3) the right caudate (18, 24, 18), and 4) the left orbitofrontal cortex (-16, 36, -22) (red pseudo-colored in B; the right caudate, located in the subcortex, was not displayed in 3-dimensional rendering). *p<0.05; **p<0.01. MADRS, Montgomery–Åsberg Depression Rating Scale; FC, functional connectivity; sgACCRS, superior part of the subgenual anterior cingulate cortex; tFUS, transcranial focused ultrasound.
As the secondary outcomes, we evaluated the changes in clinical symptoms within each group and between groups using various symptom assessment scales and neurocognitive function tests (shown in Table 2, Supplementary Tables 1 and 2). We found a significant main effect of session sequence on QIDS-SR (F1,21=22.8, p<0.001), SSI (F1,21=15.2, p<0.001), and K-POMS_total (F1,21=23.3, p<0.001) as well as a group-by-session sequence interaction effect on SSI (F1,21=5.76, p=0.015) and K-POMS_total (F1,21=10.6, p=0.001). Based on this analysis, the QIDS-SR, SSI, and K-POMS_total scores decreased over time in both groups, though the SSI and K-POMS_total scores exhibited a greater reduction in the verum group than in the sham group. Notably, a main effect of the group was only found for SSI (F1,21=5.05, p=0.035), which suggests that SSI decreased to a greater extent in the verum group than in the sham group across all sessions. No clinically significant changes were observed in the neurocognitive function tests administered through the CANTAB, except for a notable main effect of the session that reflected a learning effect.
Estimated in-situ acoustic intensity and sonication accuracy
Supplementary Figure 3 provides an example of a numerical simulation of acoustic propagation. The estimated acoustic intensity and spatial deviations in targeting during the verum stimulation session were calculated based on the simulation (Supplementary Table 3). The mean acoustic intensity at the target was 35.9%±14.0%, and the mean spatial error from the target was 2.7±0.6 mm.
fMRI data quality results
When we compare the patients between verum group with sham group, we found significant differences in invalid scans, maximum motions, maximum and mean global signal changes, between two groups (invalid scans [n], verum: 0.22±0.67 versus sham: 5.00±5.27 [p=0.02, t=2.56]; maximum motions [mm], verum: 0.58±0.15 versus sham: 1.21±0.75 [p=0.03, t=2.32]; maximum global signal changes motions [%], verum: 3.78±0.74 versus sham: 8.49±4.71 [p=0.01, t=2.81]; mean global signal changes motions [%], verum: 0.80±0.04 versus sham: 0.88±0.07 [p=0.01, t=2.93]). However, mean motions (mm) (verum: 0.13±0.03 versus sham: 0.17±0.07 [p=0.01, t=2.81]) were not different between two groups.
FC analysis
Supplementary Table 1 present statistical analysis data for subgroups of 19 individuals. The results show patterns similar to those in the analyses of all 23 subjects, which suggests that the verum group showed a greater reduction in depressive symptoms with time compared with the sham group.
The results of comparing FC with sgACC before and after stimulation in the verum group are shown in Table 3. After sonication, FC increased between the sgACCLS and bilateral medial prefrontal cortex, sgACCRS and left medial prefrontal cortex, sgACCLI and cerebellum, and sgACCLI and right middle frontal gyrus. We did not find changes in FC between any neural substrates following sham stimulation.
Brain regions showing a significant difference in FC with subregions in the sgACC before and after sonication in the verum group
Table 4, Figure 2B and C show the group difference (before and after stimulation) in the degree of FC changes with respect to the sgACCRS. Prior to stimulation, there was no significant difference in resting-state FC with the sgACCRS between groups. However, FC to the sgACCRS showed a significant degree of tFUS-mediated changes between the verum and sham condition in the left medial prefrontal cortex, left middle frontal gyrus, right caudate, and left orbitofrontal cortex. No changes in FUS-mediated FC to the stimulated left DLPFC area were observed in either of the group (within and between groups). No correlation was found between the changes in MADRS scores and FC strengths.
Brain regions showing a significant difference in functional connectivity with the sgACCRS between groups after transcranial focused ultrasound stimulation
DISCUSSION
This randomized, double-blind, sham-controlled study presents the first human clinical trial exploring the use of low-intensity tFUS stimulation to treat MDD. We found that tFUS stimulation of the left DLPFC showed a significant treatment effect, reducing suicidal ideation and depressive symptoms compared with the sham stimulation. In addition, tFUS was well tolerated by all the participants without any adverse events.
The MADRS scores of the sham stimulation group decreased over time, which we believe stemmed from the placebo effect. However, the reduction in MADRS scores was more profound in the verum group, which was evident immediately after the completion of a series of tFUS sessions; it was maintained for at least two weeks. The STAI scores also exhibited greater decreases in the verum group than in the sham group over time. These results suggest that tFUS may effectively reduce both depression and anxiety in patients with MDD.
Our results further demonstrate that tFUS may positively affect various dimensions of mood state (as assessed by K-POMS) in MDD patients. Moreover, SSI scores were significantly reduced in the verum group, which indicates that tFUS stimulation of the left DLPFC may be particularly effective in reducing suicidal thoughts. We also note that the reduced QIDS-SR scores were not significantly different between the two groups, which may be attributed to the potential bias inherent in subjective scales including personality factors. However, while acknowledging the potential bias of self-report scales, we also highlight that our study did observe some meaningful differences in other several subjective indicators. For example, significant improvements in emotional symptoms, including depression, anxiety, and other aspects of mood state (evaluated through subscales of K-POMS and STAI) in the verum group indicate the potential effectiveness of tFUS as a therapeutic intervention.
The extant research has only examined changes in FC after tFUS treatment among healthy participants, not including psychiatric populations [35]. In this study, within-group comparisons of FC in relation to tFUS treatment revealed that only the verum group showed increased connectivity between the sgACC seeds and the bilateral medial prefrontal cortex, cerebellum, and right middle frontal gyrus. This finding was not observed in the sham group, indicating that the observed enhanced connectivity was specifically related to tFUS treatment. Previous neuroimaging studies already established that patients with MDD exhibit functional brain abnormalities, including functional alterations in fronto-limbic mood regulation circuitry [36]. tFUS stimulation strengthened the FC among the brain areas in the circuitry.
Between-group FC analysis of the post-stimulation fMRI scans revealed that FUS yielded higher connectivity between the sgACCRS and several brain regions, including the left medial prefrontal cortex, left middle frontal gyrus, right caudate, and left orbitofrontal cortex. However, other subregions of the sgACC did not exhibit these between-group differences. These results are consistent with the results of previous studies [37] demonstrating that different locations within the sgACC exhibited different functional relationships with the frontal, limbic, and cerebellar regions. For example, Palomero-Gallagher et al. [38] revealed that the superior part of the sgACC seemed to have a significant association with sadness. In contrast, the inferior part of the sgACC manifested a stronger link to fear processing [38]. As tFUS has the unique ability to non-invasively stimulate deep structures in the brain, stimulation of the specific subregion of the sgACC is an attractive subject for future research.
Previous studies have reported that modulations of the cortico-limbic circuit are associated with the treatment effects of antidepressants, electroconvulsive therapy, and TMS [39,40]. Despite the distinctive changes in FC observed across the cortico-limbic circuitry, we did not find any significant correlation between symptom reduction and FC changes associated with tFUS treatment. We could also not find significant FC changes between the stimulated area (DLPFC) and the sgACC. This may be due to the small sample size, but it may imply the involvement of other neural networks related to depression rather than the cognitive control network involving regions including DLPFC and sgACC [41]. In other words, the mechanism of tFUS acting on the DLPFC may indirectly affect networks related to symptom domains such as dysphoria, rumination, and anhedonia, rather than directly altering connectivity in the cognitive control network including the connectivity between the DLPFC and sgACC. Therefore, in future confirmatory clinical trials, conducting more in-depth and comprehensive connectivity analyses using a larger sample size could be beneficial. This approach would help to explore and understand the multifaceted connectivity changes associated with tFUS treatment in different symptom domains. In addition, while the overall number of invalid scans was minimal and diligent efforts were made to correct motion artifacts, it is worth noting that the sham group exhibited more motions, which may have influenced the lack of significant outcomes. We also note that the inclusion of patients with a range of severities, rather than employing stringent patient selection criteria, might have required a more homogeneous patient group for the study to yield significant results.
When it comes to discussing the target location for stimulation, in this study, we focused on the DLPFC region, which has been extensively investigated as a most efficient target in previous neuromodulation techniques. However, for future follow-up studies, it is conceivable to explore stimulation targeting subcortical deep regions, which can further leverage the advantages of tFUS more effectively.
In this study, the mean spatial error of FUS stimulation from the target was 2.7±0.6 mm, which is far smaller than the spatial resolution offered by TMS and high-definition tDCS [5,42]. The accuracy and spatial selectivity of tFUS are advantageous; however, the technique requires more training time for the operators as it involves CT and MRI data processing and image-guided transducer positioning. We also note that none of the patients reported auditory phenomena during the stimulation. In contrast, previous studies reported that tFUS stimulation may elicit these phenomena, confounding the interpretation of the somatosensory and motor responses [43,44]. We surmise that the use of low-intensity ultrasound, along with a sonication path that is far from the auditory pathway and cranial structures, may have contributed to the rather “silent” tFUS stimulation. However, the use of higher acoustic intensity would inevitably produce auditory phenomena through bone conduction, and appropriate countermeasures, such as auditory masking with an external sound, are recommended in future studies.
Although we reported promising results of low-intensity tFUS for MDD, this study has several limitations. First, our sample size was relatively small even though we tried to investigate the patients using well-designed randomized, double-blind, sham-controlled trial. Because of this problem, it might not be possible to find a correlation between treatment effect and changes in FC. Future studies could benefit from larger sample sizes to validate the findings of the current study and explore deeper insights into treatment effects. Specifically, with a sufficient sample size, focusing on some MADRS items that are highly associated with the prefrontal cortex function could potentially enhance the analysis and yield a larger effect size. Other limitation includes the exclusion of the MDD patients with psychotic features, the patients with psychiatric comorbidities including schizophrenia and bipolar disorders. These exclusions may limit the generalization of our novel findings. Another limitation is that several patients were taking antidepressants during the study period, which could influence on the results although we did not allow medication dose change 4 weeks before and during the study. Additionally, the lack of a formal blinding evaluation is another limitation of this study. Finally, a notable omission in our study is the absence of an evaluation for hypomanic switches. Despite vigilant patient monitoring by the psychiatrists, the lack of formal assessment for hypomanic symptoms presents a gap in our assessment of potential adverse effects.
We explored a potential therapeutic application of low-intensity tFUS stimulation for treating MDD. Our study suggests that this treatment could lead to improvements in depressive symptoms and possible FC changes in patients with MDD. While this is an exploratory study, our findings provide insight into its feasibility and safety, opening doors for further neuromodulation research using tFUS in psychiatry. In future analyses of the confirmatory clinical trial with a larger sample size, we aim to replicate and refine these results.
Supplementary Materials
The Supplement is available with this article at https://doi.org/10.30773/pi.2024.0016.
Subgroup (patients with fMRI) comparison of demographic characteristics and changes in clinical symptoms between verum and sham treatment groups
Additional results of the comparison of changes in neurocognitive function and clinical symptoms between verum and sham treatment groups
The coordinates of the targeted brain area and acoustic focus obtained by numerical simulation, estimated acoustic intensity (Isppa) at the sonication focus and rate of transmission (%), and estimated spatial error in targeting during the verum stimulation session
The group-wise changes in MADRS score from the first neuropsychological assessment battery (Baseline) to the third (Follow-up) in a subgroup of patients with fMRI data. The blue line indicates the mean change in MADRS scores in the verum group (N= 9), and the orange line shows that in the sham group (N=10), with the error bars representing standard error. We found a significant main effect of session sequence (baseline, second, and third) (F1,17= 23.7, p<0.001), main effect of group (verum versus sham) (F1,17=5.8, p=0.028), and group-by-session sequence interaction effect (F1,17= 7.3, p=0.008) on the MADRS scores. In the post hoc t-tests, no difference was found in baseline (t=0.2, p=0.840), but significant group differences were found in end of treatment and follow-up (t=2.6, p=0.015 and t=2.9, p=0.009, respectively). In the verum group, we found a significant difference in MADRS scores between baseline and end of treatment (t=4.7, p=0.001) and baseline and follow-up (t=4.4, p=0.002) but no difference between end of treatment and follow-up (t=2.2, p=0.063). Similarly, in the sham group, the MADRS scores decreased between baseline and end of treatment (t=2.5, p=0.036), though no significant difference was found between baseline and follow-up (t=2.1, p=0.069) and end of treatment and follow-up (t=0.8, p=0.430). MADRS, Montgomery–Åsberg Depression Rating Scale; fMRI, functional magnetic resonance imaging.
Visual representation of individual participants’ MADRS changes from “baseline” to “follow-up” in the form of a scatter plot. V, verum; S, sham; MADRS, Montgomery–Åsberg Depression Rating Scale.
An exemplary numerical simulation of acoustic propagation from one individual overlaid on anatomical magnetic resonance imaging. Tri-planar sections are shown at the maximal acoustic intensity.
Notes
Availability of Data and Material
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
Conflicts of Interest
The authors have no potential conflicts of interest to disclose.
Author Contributions
Conceptualization: Seung-Schik Yoo, Jeong-Ho Seok. Data curation: Jin Sun Ryu, Junhyung Kim, Soojeong Kim, Hyu Seok Jeong. Formal analysis: Jooyoung Oh, Hyun-Chul Kim. Funding acquisition: Seung-Schik Yoo, Jeong-Ho Seok. Investigation: Jooyoung Oh, Jin Sun Ryu, Soojeong Kim, Kyung Ran Kim. Methodology: Seung-Schik Yoo, Jeong-Ho Seok. Project administration: Seung-Schik Yoo, Jeong-Ho Seok. Resources: Seung-Schik Yoo, Jeong-Ho Seok. Software: Jin Sun Ryu, Junhyung Kim, Soojeong Kim, Hyu Seok Jeong. Supervision: Seung-Schik Yoo, Jeong-Ho Seok. Validation: Jin Sun Ryu, Junhyung Kim, Hyu Seok Jeong, Kyung Ran Kim. Visualization: Jooyoung Oh, Seung-Schik Yoo, Jeong-Ho Seok. Writing—original draft: Jooyoung Oh. Writing—review & editing: Jooyoung Oh, Hyun-Chul Kim, Seung-Schik Yoo, Jeong-Ho Seok.
Funding Statement
This work was partially supported by Neurosona Co. Ltd., Republic of Korea, and by a Korea Medical Device Development Fund grant funded by the Ministry of Health & Welfare, Republic of Korea (Project Number: RS- 2020-KD000282).
Acknowledgements
The authors would like to thank Sun-Woo Choi, Sun-Young Mun, and Jae-Yoon Shim for their assistance in performing the tFUS treatment procedures. We also acknowledge Seul-Ah Lee and Hyun Kyung Shin for measuring psychological data and Jared Van Reet for editorial assistance.