Intermittent theta burst stimulation of cerebellar vermis enhances fronto-cerebellar resting state functional connectivity in schizophrenia with predominant negative symptoms: A randomized controlled trial

Abstract

Objective:

Negative symptoms of schizophrenia are substantially disabling and treatment resistant. Novel treatments like repetitive transcranial magnetic stimulation (TMS) need to be examined for the same using the experimental medicine approach that incorporates tests of mechanism of action in addition to clinical efficacy in trials.

Methods:

Study was a double-blind, parallel, randomized, sham-controlled trial recruiting schizophrenia with at least a moderate severity of negative symptoms. Participants were randomized to real or sham intermittent theta burst stimulation (iTBS) under MRI-guided neuro-navigation, targeting the cerebellar vermis area VII-B, at a stimulus intensity of 100% active motor threshold, two sessions/day for five days (total = 6000 pulses). Assessments were conducted at baseline (T0), day-6 (T1) and week-6 (T2) after initiation of intervention. Main outcomes were, a) Scale for the Assessment of Negative Symptoms (SANS) score (T0, T1, T2), b) fronto-cerebellar resting state functional connectivity (RSFC) (T0, T1).

Results:

Thirty participants were recruited in each arm. Negative symptoms improved in both arms (p < 0.001) but was not significantly different between the two arms (p = 0.602). RSFC significantly increased between the cerebellar vermis and the right inferior frontal gyrus (p_cluster-FWER = 0.033), right pallidum (p_cluster-FWER = 0.042) and right frontal pole (p_cluster-FWER = 0.047) in the real arm with no change in the sham arm.

Conclusion:

Cerebellar vermal iTBS engaged a target belonging to the class of cerebello-subcortical-cortical networks, implicated in negative symptoms of schizophrenia. However, this did not translate to a superior clinical efficacy. Future trials should employ enhanced midline cerebellar TMS stimulation parameters for longer durations that can potentiate and translate biological changes into clinical effects.

1. Introduction

Schizophrenia is a severely disabling medical disorder that continues to stand among the top 20 causes of disability worldwide (James et al., 2018) with a greater impact in middle-income countries (Charlson et al., 2018). Compared to delusions and hallucinations, which are more likely to remit with antipsychotic treatment (Aleman et al., 2017; Carbon and Correll, 2014; Green, 2016) negative symptoms and cognitive impairments are persistent, difficult-to-treat and strongly contribute to disability (Bhagyavathi et al., 2015; Green, 2016; Strassnig et al., 2015). The treatment of negative symptoms remains a major challenge for the field, especially since a broad range of interventions have yielded mixed results with minimal clinical utility (Fusar-Poli et al., 2015; Helfer et al., 2016; Saavedra-Velez et al., 2009; Singh and Singh, 2011; Tsapakis et al., 2015). An experimental medicine approach that combines clinical efficacy trials with tests of mechanism of action is a promising initiative that can deliver reliable and reproducible clinical benefits (Insel, 2015).

Repetitive transcranial magnetic stimulation (rTMS) alters behavior on the basis of neural network modulation and when combined with functional magnetic resonance imaging (fMRI), can be a potential neuroscience-informed treatment (Fox et al., 2014; McClintock et al., 2011; Pascual-Leone et al., 2011; Shafi et al., 2012) for persistent negative symptoms. Given the critical role of the prefrontal cortex in negative symptoms (Wolkin et al., 1992), most studies have examined the effects of high-frequency rTMS targeted to the dorsolateral prefrontal cortex (DLPFC). Although this method holds some promise for the treatment of negative symptoms, the aggregate evidence for its efficacy has been modest at best (Aleman et al., 2018; He et al., 2017; Kennedy et al., 2018; Osoegawa et al., 2018) emphasizing the necessity to explore novel target areas in the brain. Cognitive dysmetria or poor mental coordination is an integrative theory of schizophrenia characterized by the fundamental cognitive deficit of an inability to effectively coordinate information processing (Andreasen et al., 1998). This function has been pinned down to the dysfunction of cortical-subcortical-cerebellar circuitry (Andreasen et al., 1998, 1996). This dysfunction has been implicated to translate into diverse symptoms of schizophrenia including negative symptoms (Andreasen et al., 1999; Honey et al., 2005). Many recent functional connectivity studies also imply this network in the brain as a central substrate underlying negative symptoms of schizophrenia (Brady et al., 2019; Gao et al., 2019). Also given the role of the cerebellum in cognition (Allen, 1997), emotion (Buckner, 2013; Schmahmann, 2019), motivation and social behaviors (Carta et al., 2019), it could be a potentially important target for improving negative symptoms through neuromodulation. A few contemporary investigations have examined midline cerebellar stimulation (Demirtas-Tatlidede et al., 2010; Garg et al., 2016). Theta burst stimulation (TBS) is a novel patterned rTMS technique, which reduces stimulation time and causes more potent and longer-lasting effects when compared to standard rTMS (Huang et al., 2005a). A recent two-phase experiment identified a cerebellar-prefrontal network to be associated with negative symptom severity in schizophrenia; engaging this network with a five-day intermittent TBS (iTBS) treatment enhanced the connectivity in this network, which was associated with improvement in negative symptoms (Brady et al., 2019). These observations lend support to past studies identifying midline cerebellar structural changes (Heath et al., 1982) and their potential impact on regulating prefrontal activity via the striatum and thalamus as a key pathophysiological mechanism in negative symptoms of schizophrenia (Andreasen et al., 1998, 1996; Rüsch et al., 2007). The fronto-cerebellar network is a potential target network that can be engaged to improve negative symptoms in more systematic controlled trials based on an experimental medicine approach.

We therefore used a double-blind randomized controlled experiment to characterize changes in (a) severity of negative symptoms and cognitive impairment, following 10 sessions of real or sham midline cerebellar iTBS in schizophrenia patients with predominant negative symptoms, (b) fronto-cerebellar resting state functional connectivity (RSFC). We hypothesized a greater improvement in negative and cognitive symptoms and greater enhancement in the fronto-cerebellar RSFC in patients receiving real (as compared to sham) cerebellar stimulation.

2. Materials and methods

This study was a double-blind, randomized, parallel, sham-controlled superiority trial with an allocation ratio of 1:1. The trial was registered prior to recruitment of participants under Clinical Trials Registry – India (CTRI) with registration number CTRI/2017/09/009636 (ww.ctri.nic.in). The study was approved by the institutional ethics committee and all participants provided written informed consent.

2.1. Sample size calculation

The open label study that demonstrated benefits of iTBS to cerebellar midline using a similar protocol in schizophrenia demonstrated improvements in negative symptoms to the magnitude (effect size) of 0.69 (Demirtas-Tatlidede et al., 2010). Our sample of 30 in each group had 80% power at an alpha of 0.05 to detect a standardized mean difference of about 0.7 in negative symptoms between real and sham iTBS groups (Cohen, 1988).

2.2. Participant recruitment

Participants (aged 18–50 years) meeting the Diagnostic and Statistical Manual-5 (DSM-5) diagnosis of schizophrenia as evaluated by two clinicians independently, and confirmed using the Mini International Neuropsychiatric Interview (Sheehan et al., 1998), were recruited for the trial from both outpatient and inpatient services at a south Indian tertiary care center, National Institute of Mental Health and Neurosciences (NIMHANS). Each participant’s medications had to be unchanged in the last three months with no Electroconvulsive Therapy (ECT) prescribed for the last six months before recruitment. Participants needed to have a minimum required severity of negative symptoms defined as a score of three (moderate) or more on each of the five global rating items of Scale for the Assessment of Negative Symptoms (SANS) (Andreasen, 1989; Andreasen et al., 2005). To avoid confounders likely to affect cognition or treatment response, participants with clinically determined neurological disorders (e.g., seizure disorder, traumatic brain injury), substance abuse or dependence (except nicotine) within the past three months and self-reported visual or auditory impairment interfering with study assessments were excluded. We also excluded participants with persistent suicidal or homicidal behavior, pregnant, lactating or postpartum states. To minimize transcranial magnetic stimulation/TMS-related complications, participants fulfilling any of the risk factors for TMS procedures as assessed using a standard screening tool (Rossi et al., 2011) were excluded. The randomization sequence was generated through a computer by MK. UMM ensured allocation concealment by maintaining opaque sealed envelopes containing allocation status. DI/MVT opened the opaque envelopes only after the informed consent was obtained by RB from the participant thus ensuring allocation concealment. TMS was administered by DI/MVT and the assessments were performed by RB who was unaware of the treatment status. Subjective appraisals from participants and rater (RB) were recorded regarding real versus sham status of the treatment at the end of sessions to assess the effectiveness of blinding.

2.3. Clinical, cognitive and safety assessments

Negative symptoms were assessed using the SANS (Andreasen, 1989). Positive symptoms were assessed using Scale for Assessment of Positive Symptoms (SAPS) (Andreasen, 1984). Depressive and extrapyramidal symptoms were assessed by Calgary Depression Scale for Schizophrenia (CDSS) (Addington et al., 1990) and Simpson Angus Scale (SAS) (Simpson and Angus, 1970) respectively. General neurocognitive (Rao et al., 2004) and social cognitive assessments (Behere et al., 2008; Mehta et al., 2011) were performed to cover dimensions indicated in the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Cognitive Consensus Battery (Nuechterlein et al., 2004) (Appendix A). Pulse rate, blood pressure and motor cerebellar signs (International Co-operative Ataxia Rating Scale-ICARS) (Trouillas et al., 1997) were measured before and after each TMS session. Potential side effects of TMS (Rossi et al., 2009) were checked for each subject after each treatment session. All the assessments were performed thrice for each participant i.e., at baseline (T0), sixth day (T1) and sixth week (T2) after the initiation of iTBS (Supplemental Fig. 1).

2.4. TMS intervention

T1 structural magnetic resonance imaging (MRI) was obtained from all participants prior to initiation of TMS in which the cerebellar vermis (lobule VIIB) was identified by cerebellar anatomical landmarks within the proportional stereotaxic space of Talairach (Schmahmann et al., 1999). See Supplemental Fig. 2 for detailed localization procedure followed for each participant. rTMS was delivered as intermittent theta burst stimulation (iTBS). rTMS was delivered using MagPro X100 (MagVenture, Farum, Denmark). Dosing was at 100% of the active motor threshold, defined as the minimum intensity delivered to the left cerebral cortex, required to produce a motor evoked potential of >200 microvolts on >five out of ten trials from the contralateral (right) first dorsal interosseous, while the participant maintained voluntary contraction of about 20% of maximum (Rossini et al., 1994), measured using a hydraulic pinch gauge (Saehan Corp., Masan, South Korea). TMS was administered under MRI-guided neuronavigation using Brainsight frameless stereotaxic system (Rogue Research, Montreal, Canada) (Supplemental methods, “Neuronavigational procedure”) with the figure-of-8 Cool-B65 AP MagVenture coil, which could deliver either real or sham iTBS according to the codes fed into the TMS machine. The sides of the coil did not deliver real or sham stimulation in a fixed mode but were flexible to switch according to the codes fed into the TMS machine. The sham iTBS produced just a sound in a pattern like iTBS without any magnetic stimulation. The handle of the coil was held upwards along the sagittal plane. iTBS was administered as 20 trains of 2 s on and 8 s off cycle containing 3-pulse 50 Hz bursts at theta frequency (every 200 ms), for a total of 600 pulses delivered over 190 s (Huang et al., 2005b). The TMS protocol adhered to all safety guidelines and recommendations endorsed by the International Federation for Clinical Neurophysiology (Rossi et al., 2009). Participants underwent a total of ten TMS sessions administered as two sessions every day spaced 4 h apart and a total of 6000 pulses, for five days (Demirtas-Tatlidede et al., 2010).

2.5. Resting state functional magnetic resonance imaging (rsfMRI)

All scans were obtained with a 3 T Siemens Skyra MRI scanner using a 32-channel coil. The blood oxygenation level dependent (BOLD) echo-planar imaging (EPI) rsfMRI scan parameters were as follows: TR = 2000 ms; TE = 30 ms; flip angle = 78; slice thickness = 3 mm; slice order: Descending; slice number = 37; gap = 25%; matrix = 64 * 64 * 64 mm³, FOV = 192 * 192, voxel = 3.0 mm isotropic. The EPI sequences (10 min; 250 in number) were acquired in darkness and participants were asked to keep their eyes open during the session, lie as quietly as possible, and avoid falling asleep. A high-resolution structural T1-weighted MRI of 1-mm section with no inter-slice gap was also obtained to use for TMS-target localization, registration purposes during image processing and analyses. BOLD rsfMRI and T1-weighted MRI were obtained once at T0 and again within 2 days after completion of the 5-day TMS treatment course (T1). rsfMRI was analyzed on the imaging software FMRIB Software Library (FSL version-5.0.10). See Supplemental methods “rsfMRI_preprocessing” and “Quality check procedure of rsfMRI scans” for detailed description of the steps.

2.6. Seed based analysis of rsfMRI

The cerebellar vermis area VIIB was chosen as the primary seed as that was the target of stimulation and we intended to study the change in resting state connectivity of this midline cerebellum structure with the rest of the voxels in the brain following the stimulation. Vermis VIIB was chosen as the seed as per “Cerebellar Atlas with MNI152 space after normalization with FLIRT” (Diedrichsen et al., 2009). The seed was binarized to and inverse transformed from standard subject space (cleaned functional MRIs) using the matrix file available from the initial registration step. Average time series of rsfMRI BOLD-signal from individual voxels within this seed were extracted for each subject, and used as the model predictor in a general linear model (GLM) analysis to determine brain regions temporally correlated with it using fMRI Expert Analysis Tool (FSL-FEAT) (Woolrich et al., 2001). The resultant seed connectivity maps were registered back to standard MNI-152 space. Next, to determine significant within-and between-group differences, and relationship between functional brain connectivity changes and change in negative symptom severity, FSL’s randomize permutation tool was employed (Winkler et al., 2014). Clusters were determined by using threshold-free cluster enhancement (TFCE) (Smith and Nichols, 2009) and a family-wise error rate (FWER) corrected cluster significance threshold of p < 0.05 with 5000 permutations.

2.7. Statistical methodology

2.7.1. Clinical data

Baseline clinical and sociodemographic parameters were compared between the two groups using independent t-test or chi-square test. We adopted the standard intention-to-treat (ITT) analysis. Change in negative symptom severity, cognitive scores, and physical parameters over time because of intervention were examined through Mixed Models Repeated Measures (MMRM) analysis. Group and time-point of measurement were modeled as fixed effects. Clinical ratings were performed thrice on every participant contributing to a correlation between these scores within each participant. This correlation of repeated measures on the same participant was accounted for by modeling individual participants as a random effect and by specifying a covariance structure to the random effect, in this case the autoregressive first order (AR1). Maximum Likelihood (ML) method was used for estimating parameters in the model. The residuals obtained from the model for each outcome variable were tested for normality of distribution. Group * Time interaction effect in the MMRM models, which indicated differential improvement of symptoms in the groups over time, was considered statistically significant at p ≤ 0.05. All statistical analyses were performed on R Studio version 3.6.1.

2.7.2. Neuroimaging data

Baseline resting state functional connectivity maps of the cerebellar seed to the rest of the brain (seed-to-voxel) were compared between the two groups using an unpaired higher-level GLM test design within FSL. Change in functional connectivity within the two groups was separately examined using paired higher-level GLM test designs. Difference maps were created using the subtraction function in FSL by subtracting pre-from post-TMS functional connectivity maps. An unpaired higher-level test was then used to compare the difference maps between real iTBS and sham iTBS groups (with and without clozapine status as a covariate). GLM model consisting of the difference maps and the change in SANS scores was used to examine the relationship between change in whole brain functional connectivity of the cerebellar seed and change in SANS severity score following TMS.

3. Results

Thirty participants were recruited in each group from August 2016 to July 2018. The Consolidated Standards for Reporting Trials (CONSORT) diagram is depicted in Fig. 1.

Fig. 1.

CONSORT flow diagram.

Legend: Consolidated Standards for Reporting Trials (CONSORT) flow diagram depicting the recruitment of participants into the trial from screening to inclusion of data in the final statistical analysis. ITT = Intention to treat; iTBS = intermittent theta burst stimulation.

3.1. Baseline clinical and sociodemographic variables

The 2 groups were comparable on baseline, clinical and cognitive parameters except the long-term percent retention score of verbal memory which was higher in sham iTBS arm. However, a higher proportion of participants in the real iTBS arm were on clozapine (30%) compared to sham iTBS (6.67%) (p = 0.02) (Table 1, Supplemental Table 1).

Table 1.

Baseline symptom severity and cognitive scores.

	Real iTBS (N = 30)	Sham iTBS (N = 30)	t/X2	p
Scale for the Assessment of Positive Symptoms (SAPS)	4.63 (6.14)	6.10 (9.33)	−0.72	0.48
Scale for the Assessment of Negative Symptoms (SANS)	85.23 (11.75)	82.47 (13.82)	0.84	0.41
Calgary Depression Scale in Schizophrenia (CDSS)	1.27(2.59)	1.7 (2.38)	−0.68	0.50
Simpson Angus Scale (SAS)	1.2 (1.83)	0.83 (1.32)	0.89	0.38
Learning Score	31.11 (9.11)	32.33 (9.27)	−0.51	0.61
Memory Score	22.96 (7.55)	25.9 (7.44)	−1.48	0.15
Long Term Percent Retention Score	63.16 (34.05)	86.31 (38.89)	−2.41	0.019
Complex Figure Test (CFT)-copying	31.66 (5.24)	29 (29.21)	1.42	0.16
CFT-Immediate Recall	12.61 (8.75)	10.91 (8.22)	0.75	0.45
CFT-Delayed Recall	11.93 (8.34)	9.67 (7.53)	1.07	0.29
DSST (seconds) (Digit Symbol Substitution Test)	372.04 (206.9)	475.17 (288.56)	−1.53	0.13
DVT (seconds) (Digit Vigilance Test)	772.78 (373.64)	780.28 (349.25)	−0.08	0.94
Colour Trails-A (seconds)	110.48 (61.73)	162.28 (127.01)	−1.92	0.06
Colour Trails-B (seconds)	255.81 (148.14)	349.24 (221.52)	−1.82	0.075
Verbal1back Hits	6.54 (2.8)	7.03 (2.25)	−0.75	0.46
Verbal1back Errors	3.93 (2.97)	2.93 (2.78)	−0.93	0.36
Verbal2back Hits	4.18 (2.76)	4.00 (2.30)	0.27	0.79
Verbal2back Errors	6.32 (3.49)	7.2 (3.7)	−0.93	0.36
Spatial Span Total	11.64 (4.39)	10.1 (2.92)	1.59	0.12
FAS Total	17.44 (10.03)	17.52 (7.13)	−0.032	0.97
Stroop Effect Score	251.6 (122.00)	281.00 (152.66)	−0.71	0.48
First Order Theory of Mind (FOT) Index	0.78 (0.29)	0.79 (0.22)	−0.11	0.92
Second Order Theory of Mind (SOT) Index	0.34 (0.25)	0.44 (0.29)	−1.33	0.19
Faus Pax (FP) Index	0.39 (0.17)	0.39 (0.21)	−0.102	0.92
Externalizing Bias (EB) Index	−0.17 (2.8)	1.1 (2.5)	−1.65	0.11
Personalizing Bias (PB) Index	0.77 (0.34)	0.78 (0.28)	−0.06	0.96
Social Perception (SP) Index	0.68 (0.14)	0.61 (0.10)	1.00	0.32
Emotion Recognition (ER) Index	0.56 (0.16)	0.57 (0.14)	0.57	0.86

Note: All values in cells are mean (SD).

iTBS = intermittent theta burst stimulation.

p is significant at </=0.05

3.2. Mixed Models Repeated Measures analysis

The Akaike’s Information Criterion (AIC) was the least for the AR1 covariance structure type compared to other models confirming a better fit of the AR1 model. There was no significant difference between the two arms in improvement of negative symptoms with Group * Time interaction, F(df) = 0.511(2,110.98), p = 0.602 (Fig. 2, Table 2). The two interventions did not differ significantly from each other on change in other clinical, cognitive, and physical measures (Table 2, Supplemental Tables 2 & 3). However, both the groups showed improvement in negative symptoms, cognitive parameters (except the long-term retention score, Color Trails-A, Verbal 1Back hits, Verbal 2Back errors and the social cognitive variables) over time (Table 2, Supplemental Table 2). Among the physical measures, the diastolic blood pressure, extrapyramidal symptoms score, and the ataxia score decreased over time in both the arms (Supplemental Table 3). Change in global scores of individual symptom domains of the SAPS and SANS was also not significantly different between the two groups (Supplemental Table 4). All the residuals obtained from the linear mixed models satisfied normal distribution. As the real iTBS arm had more treatment resistant participants on clozapine (failure to respond to two adequate trials of antipsychotics), which in turn could confound the response to the intervention despite the similarity in symptom severity at baseline, Group * Time * Clozapine interaction was studied, which was not significantly different for negative symptom severity between the two groups (p = 0.423). The process of blinding was effective as there was no difference between the two groups in the subjective appraisal from participants and the rater, regarding the arm of treatment to which they were assigned (Supplemental Table 5).

Fig. 2.

Comparison of change in negative symptom severity between the 2 groups.

Legend: A: Scatter+Box+Violin plots showing the change in severity of negative symptoms across time-points in the 2 groups with no significant difference (Group * Time interaction effect, p = 0.602). The notch of the inset boxplot corresponds to the median, the red dot represents the mean, the whiskers show the first and the third quartiles and the points represent the individual scores. B & C: Percentage of change in severity of negative symptoms of individual participants belonging to the two groups. SANS=Scale for the Assessment of Negative Symptoms, iTBS = intermittent theta burst stimulation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

$$\text{Percentage of change}=\frac{\text{Baseline SANS}-\text{Day}-6/\text{Week}-6\text{SANS}}{\text{Baseline SANS}}*100$$

Table 2.

Effect of real iTBS v/s sham iTBS on symptom severity.

Measure	Timepoint	Real iTBSa Mean (SE)	Sham iTBSa Mean (SE)	F(df) (Group effect)	p (Group effect)	F(df) (Time effect)	p (Time effect)	F(df) (Group * Time effect)	p (Group * Time effect)
Scale for the Assessment of Negative Symptoms (SANS)	Baseline	86.35 (2.5)	84.89 (2.93)	1.144 (1,62.39)	0.289	31.604 (2,110.99)	<0.001	0.511 (2,110.98)	0.602
	Day6	77.69 (2.5)	73.72 (2.93)
	Week6	76.4 (2.7)	71.97 (3.0)
Scale for the Assessment of Positive Symptoms (SAPS)	Baseline	5.24 (1.36)	7.41 (1.6)	1.246 (1,57.10)	0.269	9.675 (2,105.74)	<0.001	0.082 (2,105.73)	0.921
	Day6	2.64 (1.36)	4.69 (1.6)
	Week6	4.7 (1.45)	6.2 (1.63)
Calgary Depression Scale in Schizophrenia (CDSS)	Baseline	1.17 (0.43)	1.49 (0.49)	0.074 (1,58.79)	0.787	4.54 (2,106.34)	0.013	1.913 (2,106.33)	0.153
	Day6	0.34 (0.43)	0.59 (0.49)
	Week6	1.44 (0.48)	0.48 (0.51)

df = degrees of freedom, SE = standard error.

iTBS = intermittent theta burst stimulation.

p is significant at </=0.05

^aEstimated marginal means reported.

3.3. rsfMRI

Twenty-five participants in the real iTBS group and 30 in the sham iTBS group had good quality resting state fMRI data available for analysis. GLM comparing the cerebellar vermis area VIIB seed–whole brain connectivity maps between the two groups at baseline (pre-intervention) did not show any statistically significant differences. GLM comparing the connectivity maps before the intervention to that post-intervention in the real iTBS arm showed significantly increased connectivity between the cerebellar vermis area VIIB and areas corresponding to the right inferior frontal gyrus (IFG) (a threshold free enhancement family wise error rate corrected p value (p_cluster-FWER) of 0.033, with a size of 278 voxels, with peak MNI coordinates of 44,16, 20), right pallidum (p_cluster-FWER = 0.042, size = 42 voxels, peak MNI coordinates of 20, - 12, 0) and right frontal pole (p_cluster-FWER = 0.047, size =17 voxels, peak MNI coordinates of 6, 62, 2) (Fig. 3). There was no significant difference observed between the pre and post intervention connectivity maps of the sham iTBS group (p_cluster-FWER = 0.179). GLM comparing the subtraction maps (post intervention – pre intervention) of the two groups revealed no significant differences (p_cluster-FWER = 0.207). Given that the real iTBS arm had more treatment resistant participants on clozapine, addition of clozapine status as a covariate in the GLM comparing the subtraction maps did not yield any significant difference (p_cluster-FWER = 0.126) either. There was no significant association between change in connectivity of the cerebellar area VIIB after TMS and change in negative symptom severity in the whole group (p_cluster-FWER = 0.077) or individually in either of the groups (Real iTBS: pcluster-FWER = 0.275, Sham iTBS: pcluster-FWER = 0.286) (Table 3). The baseline functional connectivity maps of the whole group, real and sham iTBS arms are illustrated in Supplemental Fig. 3.

Fig. 3.

Change in resting state functional connectivity in real iTBS arm.

Legend: Sagittal, coronal, and axial (from left to right in each panel) sections of the brain demonstrating increased resting state functional connectivity between cerebellar seed, vermal area VIIB and A. Right inferior frontal gyrus (p_cluster-FWER = 0.033), B. right pallidum (p_cluster-FWER = 0.042) and C. right frontal pole (p_cluster-FWER = 0.047) from baseline to post-intervention in the real iTBS arm. iTBS = intermittent theta burst stimulation, p_cluster-FWER = threshold-free cluster enhancement and a family-wise error rate (FWER) corrected cluster significance.

Table 3.

rsfMRI results of different contrasts showing the peak MNI co-ordinates of clusters with the lowest p-value.

Analysis	Contrast	pcluster-FWER	Cluster size (no. of voxels)	MNI coordinates
				X	Y	Z
Difference in connectivity in Real iTBS arm due to intervention	Post-pre	0.033	278	44	16	20
		0.042	42	20	−12	0
		0.047	17	6	62	2
	Pre-post	0.962	3	26	67	17
Difference in connectivity in Sham iTBS arm due to intervention	Post-pre	0.179	15	45	48	35
	Pre-post	0.647	01	56	15	49
Difference of subtraction connectivity maps (post-pre) of the two groups	Real-sham	0.207	15	39	31	51
	Sham-real	0.849	25	45	48	36
Difference of subtraction connectivity maps (post-pre) of the two groups (with clozapine status as a covariate)	Real-sham	0.126	12	40	31	51
	Sham-real	0.905	1	77	49	22
Relationship between change in connectivity and change in SANS total (whole group)	Direct	0.077	23	69	37	29
		0.077	16	72	33	36
		0.077	09	51	60	31
	Inverse	0.975	3	25	62	21
Relationship between change in connectivity and change in SANS total (real iTBS)	Direct	0.684	1	39	65	20
	Inverse	0.275	7	51	77	42
Relationship between change in connectivity and change in SANS total (sham iTBS)	Direct	0.286	1	42	22	16
	Inverse	0.955	3	40	58	26

MNI = Montreal Neurological Institute, FWER = family wise error rate, iTBS = intermittent theta burst stimulation, SANS=Scale for Assessment of Negative Symptoms.

p is significant at </=0.05

3.4. Adverse events

Two participants who received real iTBS showed symptoms suggestive of mania/hypomania during the trial whereas none who received sham iTBS showed such symptoms, a report of which has been described elsewhere (Basavaraju et al., 2020). One participant from the real arm reported neck muscle contraction and ensuing tolerable neck pain/discomfort during the TMS sessions, none was reported in the sham arm. No other adverse events were reported.

4. Discussion

In this randomized controlled experimental study, we report increased functional connectivity of the stimulated midline cerebellum with the right inferior frontal gyrus and pallidum in schizophrenia patients with predominant negative symptoms in the real iTBS arm. Both the groups showed significant improvement, with no between-group differences in the primary clinical outcome measure (SANS scores), suggesting a placebo effect. The enhanced functional connectivity within the fronto-pallidal-cerebellar network did not relate to change in negative symptoms.

The dysfunction of cortical-subcortical-cerebellar circuitry has been implicated to underlie a poor mental coordination or cognitive dysmetria which is proposed to translate into diverse symptoms of schizophrenia (Andreasen et al., 1998, 1996) including negative symptoms (Andreasen et al., 1999; Honey et al., 2005). Many recent large-scale functional connectivity studies also imply this structurally polysynaptic network in the brain as a central substrate underlying key symptoms of schizophrenia (Brady et al., 2019; Chen et al., 2013; Collin et al., 2011; Gao et al., 2019; Guo et al., 2018, 2015; Ji et al., 2019; Lee et al., 2019; Liu et al., 2011; Zhuo et al., 2018). Hence, we utilized this network as a biological target for engagement through an intervention. Target engagement is a method of approach to clinical trials called experimental medicine which has been proposed by the National Institute of Mental Health (NIMH) to search for novel targets and increase Phase III success instead of trials designed to just look for an efficacy signal (Insel, 2015). Brady et al. (2019) adopted a similar approach in a smaller sample and demonstrated that enhancing the network connectivity between midline cerebellum and the right DLPFC through iTBS correlated with improvement in negative symptoms. Our study had a TMS stimulation protocol similar to that of Brady et al. (2019). In our study, cerebellar vermal iTBS enhanced the RSFC in a network comprising the vermis area VIIB, right pallidum, right inferior frontal gyrus and the right frontal pole. Previous studies have shown abnormalities in RSFC between cerebellum and IFG in schizophrenia (Collin et al., 2011; Ji et al., 2019; Tarcijonas et al., 2019). High functioning patients with first episode schizophrenia showed increased functional connectivity of the right pallidum and right IFG compared to low functioning patients (Tarcijonas et al., 2019). Our results represent a novel right-sided tripartite cerebello-subcortical-cortical circuit being activated in schizophrenia with predominant negative symptoms through TMS.

Though we have demonstrated the engagement of a novel network and enhancement of the network’s intrinsic RSFC through vermal iTBS, there was no significant difference between the real and sham arms in improvement of negative and cognitive symptoms. Also, a lack of relationship between change in connectivity of this network and change in negative symptom severity might imply that this network may not be a suitable biological target for engagement to successfully improve negative symptoms but just a mechanism of action of vermal iTBS (Insel, 2015). Alternatively, this trial has many unique participant, trial-design and treatment-protocol related characteristics that might have predisposed to the dissociation between the imaging and the clinical results. In contrast to most other studies testing interventions for negative symptoms (Brady et al., 2019), our trial had a minimum cut-off of symptom severity for inclusion as a score of three or more on each of the five global rating items of SANS (moderate severity) and also, they had to be on stable medications for three months before recruitment in to the trial. Hence our patient group was moderate to severely ill, with predominant negative symptoms and minimal to absent positive symptoms due to stabilized antipsychotic medication (mean total SANS score above 80 and mean total SAPS score around 5), absent to minimal depressive and extrapyramidal symptoms and a longer duration of illness (>8 years). This probably represents a population of patients with enduring primary negative symptoms or deficit syndrome (Carpenter, 1994), with unique biological underpinnings (Voineskos et al., 2013; Wheeler et al., 2015), suggestive of a possible subtype of schizophrenia. On the other hand, in this trial, negative symptoms and cognitive parameters improved over time irrespective of the interventional arm which points towards a high placebo response being one of the reasons contributing to the absence of efficacy. High placebo response in a trial like ours could be due to the general trend of increase in placebo response in trials of psychiatry (Razza et al., 2018; Rutherford et al., 2014; Rutherford and Roose, 2013; Weimer et al., 2015) and schizophrenia (Gopalakrishnan et al., 2020). TMS trials additionally offer a unique challenge in accounting for the placebo effect compared to medication trials. Due to large embedded placebo effects in sham medical devices compared to pill placebo, an efficacy paradox is created, requiring RCTs that are powered higher than regular trials (Burke et al., 2019). Advancement over years in the methodological effectiveness of blinding in rTMS trials (Dollfus et al., 2016; Razza et al., 2018), improvement in rTMS delivery technology like neuronavigation (Dollfus et al., 2016), parallel study over cross-over designs (Dollfus et al., 2016) and placebo-controlled studies over active drug-drug comparator trials (Rutherford et al., 2014; Rutherford and Roose, 2013), all contribute to this phenomenon. The improvement of symptoms despite being severe and treatment-resistant can be due to possible rater bias in baseline inflation of severity which is known to happen in questionnaire-based RCTs in psychiatry (Rutherford and Roose, 2013), enriched therapeutic milieus which these withdrawn and isolated patients are exposed to in the trial compared to treatment as usual (Rutherford and Roose, 2013), and the short duration of the trial with closely spaced assessments which can beget abrupt and transient placebo responses compared to real clinical efficacy that usually has a gradual onset and is persistent (Quitkin et al., 1991). Placebo lead in periods, cross over designs, comparing cerebellar vermal iTBS with an active intervention like DLPFC stimulation or transcranial direct current stimulation (tDCS), matched in terms of durations of treatment and physician-patient contact, centralized rating blind to study inclusion severity-criteria, longer trials with adequately spaced-out assessments are some of the methods we propose to handle the issue of increased placebo response in future similar trials.

Despite a difficult-to-treat severely ill population and also with a higher proportion of treatment-resistant patients on clozapine in the real iTBS group, we have demonstrated modulation of a target network by vermal iTBS pointing towards the ability of this treatment modality in causing a significant neurobiological change in the brain. We observed that two patients belonging to the real iTBS arm showed symptoms of hypomania/mania whereas there were none in the sham arm (Basavaraju et al., 2020). In keeping with these observations, and also given the absence of physical adverse effects, we propose alteration of TMS delivery parameters like the number of sessions, number of pulses per session, and intensity of stimulation to generate improvement beyond the observed placebo response in future trials. The total pulses administered in depression trials with theta burst stimulation is between 60,000 and 90,000(Cole et al., 2020; Williams et al., 2018) compared to just 6000 total pulses in our trial. There exists evidence for a dose-response relationship between the number of pulses of TBS and the strength of functional connectivity of the modulated network (Nettekoven et al., 2014). This potentially implies a direct translation of dose of intervention to magnitudes of biological change and clinical effects yielding a promise of better clinical effects with higher doses.

One of the limitations of the study is that there was no difference in the subtraction connectivity maps between the two groups despite significant difference in modulation of connectivity within individual groups. However, we do emphasize on the within group significant change in functional connectivity of the vermis seed in the real group, since this finding partially replicates observations of Brady et al. (2019) regarding our a priori proposed target engagement. This study used a figure-of-8 coil for the cerebellar vermis which is >3 cm (Hurtado-Puerto et al., 2020) from the scalp surface. In a previous electrical field modeling study the figure-of-8 coil has been demonstrated to induce a supra-threshold electric field at midline cerebellum (Bijsterbosch et al., 2012). It is possible that targeting this deep-seated structure with sub-threshold intensity from a figure-of-8 coil might not have ensured adequate stimulation of the cerebellar vermis, which is difficult to conclude without modeling the electrical field generated in the target tissue. A double-cone coil which has a deeper penetration and hence more suitable for midline cerebellar stimulation (Fernandez et al., 2020) with concomitant electrical field modeling can mitigate this shortcoming.

This evidence of an apparent biological effect in the form of target-engagement of a cortico-subcortical-cerebellar network in a more severely symptomatic population, albeit with null clinical efficacy with the relatively lower stimulation parameters should encourage future studies to engage this target with a high-dose stimulation protocol. Thus, modulation of this newly discovered biological substrate requires replication in experiments with modifications of the various trial parameters as discussed. This would very likely ensure adequate magnetic stimulation of the brain area and aid in translating the target engagement to actual clinical efficacy. Our findings open avenues for exploring promising novel therapeutic modalities by testing the mechanism of action, through target engagement by brain stimulation techniques, for treatment of disabling symptoms of schizophrenia.

Appendix A.

Cognitive tests

Domain	Cognitive test	Task description
Neurocognition (Rao et al, n.d.)
Verbal Learning & Memory	Auditory Verbal Leaning Test	A 15-word list (A) is presented 5 times verbally followed by a recall. Then, an interference list (B) of 15 words is presented before a recall of list-A. Delayed recall and recognition are also tested
Visual Learning & Memory	Complex Figure Test	Participants are asked to draw a complex diagram, initially by copying and then are asked to redraw from memory immediately and at a delayed timepoint
Processing Speed	Digit Symbol Substitution Test	The participant is asked to fill up the blanks below numbers with corresponding symbols according to a key provided.
Attention and Concentration	Digit Vigilance Test	Participants are asked to strike out particular digits (6&9) on a sheet of paper with multiple lines of single digits.
Verbal Working Memory	N-Back Test	Participants are presented a sequence of verbal stimuli. For each stimulus the participant has to decide whether the current stimulus matches the stimulus presented N number of trials back
Visuospatial Working Memory	Spatial Span Test	Participants are presented with a spatial sequence of blocks arranged on a board with serially increasing complexity. They are expected to reproduce the spatial sequence as demonstrated.
Executive Functions	Color Trails-A&B	Participants are asked to join numbers in sequence (A) and asked to alternate between colors of the circles in which the numbers are printed (B). For example, red 1 followed by a yellow 2, red 3 and so on.
	Stroop Test	Participants are expected to read a list of words written in incongruent colors followed by mentioning the color of the ink instead of reading the word.
Social cognition
Theory of Mind/Social perception (Mehta et al., 2011)	First and second order Theory of Mind, Faux Pas, Attribution Bias and Social Cue Perception Test	Story-based tasks in which the participants are required to decipher mental states of others at different complexity levels. Tasks based on videos demonstrating social interaction between characters, where participants are required to pick up the social cues
Emotion Recognition (Behere et al., 2008)	Facial emotion recognition test	Static images and dynamic videos of six basic human emotions (happy, anger, sad, fear, surprise and disgust) depicted by trained actors