Randomized, sham-controlled trial of real-time fMRI neurofeedback for tics in adolescents with Tourette Syndrome

Abstract

Background:

Activity in the supplementary motor area (SMA) has been associated with tics in Tourette Syndrome (TS). The aim of this study was to test a novel intervention – real-time functional magnetic resonance imaging (rt-fMRI) neurofeedback from SMA – for reduction of tics in adolescents with TS.

Methods:

Twenty-one adolescents with TS were enrolled in a double-blind, randomized, sham-controlled, crossover study involving two sessions of neurofeedback from their SMA. The primary outcome measure of tic severity was the Yale Global Tic Symptom Severity Scale (YGTSS) administered by an independent evaluator before and after each arm. The secondary outcome was control over the SMA assessed in neuroimaging scans where subjects were cued to increase/decrease activity in SMA without receiving feedback.

Results:

All 21 subjects completed both arms of the study and all assessments. Participants had significantly greater reduction of tics on the YGTSS scale after neurofeedback as compared to sham (p<0.05). Mean Total YGTSS score decreased from 25.2±4.6 at baseline to 19.9±5.7 at endpoint in neurofeedback and from 24.8±8.1 to 23.3±8.5 in sham. The 3.8-point difference is clinically meaningful and corresponds to an effect size of 0.59. However, there were no differences in changes on the secondary measure of control over the SMA.

Conclusions:

This first randomized controlled trial of rt-fMRI neurofeedback in adolescents with TS suggests that this neurofeedback intervention may be helpful for improving tic symptoms. However, no effects were found in terms of change in control over the SMA, the hypothesized mechanism of action.

INTRODUCTION

Tourette Syndrome (TS) is characterized by childhood onset of motor and vocal tics affecting an estimated 14/1000 children.(1) Tics usually begin between ages 5–7 years and reach peak severity between ages 9–12 years, with gradual improvement through adolescence; a minority of cases have moderate to severe tics into adulthood.(2) Available treatments for tics include comprehensive behavior therapy for tics (3, 4) and medications.(5, 6) However, a significant proportion of patients with TS may not respond to currently available interventions. We tested a novel intervention – real-time functional magnetic resonance imaging (rt-fMRI) neurofeedback – aimed at non-invasive altering of brain function

Contingent reward for tic-free intervals was associated with tic reduction in TS (7) supporting the promise of feedback learning based interventions for TS. Rt-fMRI neurofeedback (NF) can potentially harness feedback learning to train targeted control over the neural circuitry underlying tic generation. This can complement the current frontline treatment options that rely on conscious behavioral strategies or medications to change brain function in a less localized manner.

Here we report results from one of the first registered clinical trials for rt-fMRI-NF. Other early trials of this recently developed technique (8–15) have explored its clinical utility across a range of conditions, from depression (10, 12) to Parkinson’s disease.(15) This is the first study of rt-fMRI-NF for TS.

As a neuroscience-based intervention, rt-fMRI-NF requires a model of the neural circuitry underlying symptoms. For TS, evidence supports a model of dysfunction in motor corticostriatal-thalamo-cortical loops.(16) Specifically, the supplementary motor area (SMA) appears to be a key node in the dysfunctional neural circuit underlying the chronic tics of TS. Stimulation of the SMA can produce both movements and urges to move that are strikingly similar to tics and the premonitory urges experienced by patients with TS.(17) The movements (or urges) induced by SMA stimulation vary in complexity from simple motor acts to complex movements, just as the tics experienced by patients with TS can vary from simple to complex motor acts.(17) Furthermore, a number of recent human neuroimaging studies have highlighted the role of the SMA in chronic tics(18–23) as well as in other movement disorders (15, 32,33).

In summary, there is substantial data suggesting that the SMA plays a critical role in the dysfunctional neural dynamics that give rise to chronic tics. Therefore, we conducted a rt-fMRI neurofeedback study to determine if training patients to control this region can be helpful for improving tic symptoms.

MATERIALS AND METHODS

Design

This was a randomized, double-blind, sham-controlled, crossover trial of rt-fMRI neurofeedback (NF) versus a sham feedback control condition. Subjects were randomized with 1:1 allocation to either a NF followed by sham sequence or a sham followed by NF sequence. Clinical outcomes were assessed by an independent evaluator, who was unaware of the randomization sequence and subjects were also unaware of whether they were receiving NF or sham. CONSORT flow diagram of study design is shown in Figure S5 of the Supplemental Information (SI).

Subjects

We enrolled individuals with TS ages 11 to 19 years who had a minimum Yale Global Tic Severity Scale (YGTSS)(24) Total Tic score of ≥13. Participants on psychotropic medication were eligible if the medication was stable for at least one month before baseline with no planned changes during the trial. Individuals with IQ<80, current diagnosis of substance abuse or dependence, lifetime diagnosis of autism spectrum disorder, bipolar or psychotic disorder were excluded. Those with co-occurring diagnoses (e.g., depression, anxiety, OCD or ADHD) were eligible if the condition did not require immediate treatment or change in existing treatment. Additional inclusion criteria were ability to keep head still during fMRI, activation of the SMA in the localizer task (so that we were able to identify the target area), absence of braces or ferrous medical implants, and absence of neurological conditions that can interfere with interpretation of fMRI data.

Baseline Assessments

These included the Schedule for Affective Disorders and Schizophrenia for School-Age Children (K-SADS)(25) and the YGTSS(24). Diagnosis of TS was confirmed based on the presence of both motor and vocal tics for more than one year. Intelligence was evaluated with the Kaufman Brief Intelligence Test (K-BIT).(26) Medical and treatment history and demographic data were collected by semi-structured interview.(27) The clinical interviews were conducted by an experienced independent evaluator (IE) who was not involved in any part of the intervention delivery and was blind to treatment assignment. Before each clinical outcome assessment subjects and their parents were reminded not to discuss their treatment with the IE.

Outcome measures

YGTSS(24) is a semi-structured interview of tic severity that has emerged as gold standard for measuring outcomes in clinical trials in TS.(28) The YGTSS Total Tic score (range 0 to 50) was the primary outcome measure. Another clinical outcome measure was the Clinical Global Impression-Improvement (CGI-I) scale designed to rate overall clinical change in tics.(29) Scores on the CGI-I range from very much improved (1) to very much worse (7); scores of much improved (1) or very much improved (2) were used to define positive response. The YGTSS and CGI-I were administered pre-intervention (prior to NF or sham arm) and post-intervention (approximately half a week after the NF or sham arm).

Control over the target brain area – the secondary outcome measure – was assessed in functional MRI runs referred to as “control task runs”. As in the neurofeedback runs, subjects were cued to alternately raise and lower activity in the SMA. Unlike the neurofeedback runs, however, no feedback was provided during the control task runs. These data were collected approximately half a week pre-intervention and half a week post-intervention for each arm.

For 14 subjects, the cross-over arm was scheduled about a week after completion of the first arm, and the post-treatment assessment of the first arm served as a pre-treatment assessment of the second arm. Seven subjects who had more than a week period between the first and the second arms of the study were asked to repeat outcome assessments (i.e., the YGTSS and the control task scans) 3 days before starting the cross-over phase intervention.

Standard Protocol Approvals, Registrations, and Patient Consents

The study was approved by the Institutional Review Boards. Young adult participants and parents of child participants provided consent; children provided assent. The study was registered on clinicaltrials.gov before the start of enrollment ( NCT01702077).

Randomization

Subjects who met eligibility criteria were randomized. A research team member, who was not involved in the outcome assessment, assigned subjects according to the randomization schedule. Subjects were told there was an experimental intervention (that was being testing for clinical efficacy) and a control intervention (that was not believed to have any potential for improving symptoms) and that everyone would receive each intervention, but that they would not be informed about the order. The term “sham” was not used and the nature of the control condition was not described.

Cognitive strategy discussion

Prior to beginning the study, the function of the SMA was described to the subjects and they were asked to think about specific strategies for controlling the region (see Supplementary Information (SI) for details). Subjects were told these strategies could be helpful, but that they should do whatever they found effective and should not feel obliged to adhere to specific strategies.

Neuroimaging procedures

See SI for detailed information.

Overall design

Assessments involved both clinical and neuroimaging assessments. They were scheduled prior to and half a week after each arm of feedback. Clinical assessments occurred at the Yale Child Study Center; all scans occurred at the Yale Magnetic Resonance Research Center. Each arm of feedback involved two sessions of feedback training, spaced half a week apart. So, all participants came in for seven consecutive scanning sessions (or eight if the midpoint was repeated prior to the second arm), spaced at half week intervals. Information regarding adverse events was collected at each clinical assessment visit; no serious or unintended adverse events were reported.

Localizing the target brain area

For three to six functional imaging runs at the end of the first assessment scan, subjects were instructed to imitate their most common and troublesome tics (excluding those that involved head movement) during blocks of time (cued by the word “imitate”) and to rest in alternate blocks (cued by the word “rest”). These data were analyzed with voxel-wise t-tests comparing imitate and rest blocks to allow identification of the region of their SMA controlling tic-relevant musculature in an individualized manner. Figure 1 shows the overlap of the target areas used for neurofeedback across subjects.

Figure 1.

Overlap of target regions for neurofeedback. All subjects received neurofeedback from an individually localized target region in the supplementary motor area/Brodmann area 6. This figure shows the overlap of all target regions on a template brain using radiological convention (left is on the right). Warmer colors indicate voxels that were included in a greater number of individual targets regions (from a minimum of 1 to a maximum of 13).

Neurofeedback intervention

Subjects were cued by an arrow at the top of the screen to increase (red arrow pointing up) or decrease (blue arrow pointing down) activity in their SMA. A color-coded line graph (red during increase blocks, blue during decrease blocks) at the bottom of the screen provided feedback. Real-time analysis was done as described in our previous publication.(30)

Sham feedback condition

The instructions and cues were identical to the NF condition, but the feedback provided at the bottom of the screen was taken from the NF runs of the previous subject. This form of sham (yoked sham) is designed to control for perception of success across the NF and sham conditions: each subject’s sham runs indicate the same level of success, in terms of raising the line in the red blocks and decreasing it in the blue blocks, as the previous subject’s NF runs, thereby ensuring that levels of positive feedback are similar across the two conditions.

Control task scans

These were collected in the assessment scans that were scheduled half a week before and half a week after each arm of feedback and were identical to feedback scans, except that no feedback was provided. Data were analyzed to evaluate success in controlling the SMA (i.e., raising signal in the region in red periods and lowering it in the blue periods).

Offline analysis of feedback scans

Although it was not a registered outcome measure, we examined control over the SMA and whole brain activation during the real-NF and sham feedback runs, as an exploratory index of target engagement.

Debriefing

After completing each arm of the study, subjects were asked if they believed they had just received the experimental or control intervention.

Data analysis of outcome measures

The effect of neurofeedback on the YGTSS Total Tic score was tested with a mixed-effects model which included treatment (NF or sham) and time (baseline or endpoint within each arm) as within-subjects effects, cross-over sequence (real-first-followed-by-sham or sham-first-followed-by-real) as a between-subjects effect, and random subject effects. The best fitting variance-covariance structure was selected using Schwartz’ Bayesian information criterion. The primary hypothesis of the effects of NF versus sham on tic symptoms was tested by the treatment by time interactions.

This study was conducted using a cross-over design as described above. Our group has recently identified an unanticipated pattern of symptom change that unfolds over the weeks following neurofeedback.(31) Because this pattern of continued change was not anticipated in the planning stages of this study, this study was not designed to detect or account for a carryover effect. Therefore, we examined the data carefully for evidence of carry-over effects. Results did suggest that carryover may have contaminated the second arm of the intervention. Subsequent analyses were thus restricted to the first arm of the intervention. For analyses restricted to the first arm of the study, we used a mixed-model which included treatment (NF or sham) as a between subject factor and time (baseline or endpoint) as a within-subjects factor. This analysis was used to evaluate the effect of NF on the YGTSS tic score (our primary outcome measure). An identical analysis was repeated on our secondary outcome measure of control over the SMA. In addition, Pearson correlations were computed between changes in control over the SMA and changes in YGTSS during neurofeedback.

To compare the results of this trial with those of previous studies, the effect size for the YGTSS Total Tic score was calculated for each of two arms as the mean change in the measure from baseline to endpoint during NF minus the mean change during sham divided by the SD for the whole sample at baseline for the respective arm. Categorical data, including data used to characterize subjects at baseline, the rates of positive response on the CGI-I, and self-reported perception of the intervention received were evaluated by chi-square (Fisher’s exact test when cell sizes were less than 5).

For all tests, the two-tailed statistical significance was set at alpha=0.05.

RESULTS

Baseline characteristics

Between December 2012 and October 2017, 21 subjects met the study inclusion criteria and were randomized. Of 44 subjects screened, nine were ineligible, and 14 declined to participate because of scheduling difficulties or concerns about the subject’s ability to attend regular study visits (see Figure S5). Subjects (17 male, 4 female) ranged in age from 11.6 to 20.2 years (mean ± SD age = 16.0±2.9 years). Thirteen of 21 subjects had one or more comorbid psychiatric disorders and 16 subjects were receiving psychiatric medications. There were no differences between the two groups on the level of tic severity before treatment or on any clinical and demographic characteristics at baseline (see Table 1).

Table 1.

Demographic and Clinical Characteristics by Treatment Group at Baseline.

	NF-Sham (N=11)		Sham-NF (N=10)
					Test	p-value
	Mean	(SD)	Mean	(SD)
Age	15.82	(2.55)	16.30	(3.34)	t=0.374	0.713
Full Scale IQ	106.27	(14.60)	111.63	(5.53)	t =0.981	0.340
YGTSS Total Tic score1	25.25	(4.60)	24.80	(8.12)	t =0.134	0.894
Socioeconomic status
	N	%	N	%
Male gender	9	81.8	8	80.0	Fisher Exact	1.0
Race	11	100	10	100	Fisher Exact	1.0
White	10	90.9	9	90.0	Fisher Exact	1.0
Black
Asian
Ethnicity	11	100	10	100	Fisher Exact	1.00
Hispanic	2	18.2	1	10.0	Fisher Exact	1.00
Non-Hispanic	9	81.8	9	90.0	Fisher Exact	1.00
Single-parent family
Special education services
Medication status
No medication3	2	18.2	3	30.0	Fisher Exact	0.635
Receiving medication4	9	81.8	7	70.0	Fisher Exact	0.635
α-Agonist	3	27.3	6	60.0	Fisher Exact	0.198
Antipsychotic	4	36.4	1	10.0	Fisher Exact	0.311
SSRI	2	18.2	1	10.0	Fisher Exact	1.000
Stimulant	1	9.1	2	20.0	Fisher Exact	0.586
Other	2	18.2	2	20.0	Fisher Exact	1.000
Additional diagnoses5	6	54.5	7	70.0	Fisher Exact	0.659
ADHD	2	18.2	4	40.0	Fisher Exact	0.361
OCD	2	18.2	3	30.0	Fisher Exact	0.635
Anxiety disorder	2	18.2	3	30.0	Fisher Exact	0.635
Oppositional Defiant Disorder	2	18.2	1	10.0	Fisher Exact	1.0

Analysis of full dataset (both arms)

Mixed model analysis of the YGTSS score revealed a significant main effect of time (F_1,57 = 36.56; p<0.0001) and, consistent with our primary hypothesis, a significant treatment by time interaction (F_1,57 = 4.53; p<0.05). Post-hoc analyses exploring the basis of the treatment by time interaction across both arms of the study revealed a significant decrease (mean: −4.9, 95% CI: −6.6, −3.2) in symptoms during NF (F_1,57 = 33.42, p<0.0001) that exceeded that in the sham (mean: −2.3, 95% CI: −4.0, −0.6) condition (F_1,57 = 7.67, p<0.0076). These effects for both arms are shown in Figure 2B, first red and blue bars representing mean change scores for NF and sham conditions, respectively, averaged across both arms of the study. The main effects of treatment and sequence as well as the interaction effects of sequence by time were not significant. However, there was a significant treatment by sequence interaction (F_1,57 = 53.0; p<0.0001), suggesting a possible positive carryover effect.

Figure 2.

(A) Yale Global Tic Severity Scale (YGTSS) Total Tic score at baseline and endpoint by treatment condition (Real Neurofeedback [NF], Sham Control) and study arm (Arm 1=before crossover, Arm 2=after crossover). The group that was randomized to receive Real NF first (real-first-followed-by-sham) is shown with a solid line. The group that was randomized to receive Sham Control first (sham-first-followed-by-real) is shown with a dashed line. Error bars represent standard error. (B) Change in YGTSS Total Tic score by treatment condition (Real NF, Sham Control) and study arm (Arm 1=before crossover, Arm 2=after crossover, All Data=Arm 1 and Arm 2 combined). Change scores were calculated by subtracting pre-intervention scores from post-intervention scores. A more negative value reflects greater clinical improvement from pre to post. Error bars represent standard error.

Carry-over effects

Given the significant treatment by sequence interaction and our recent finding that NF can result in symptom changes that persist for weeks following the intervention,(31) we examined the pattern of symptom improvements seen in the two sequences for evidence of carryover effects across arms. These data are summarized in Table 2 and shown in Figure 2; F-values below represent tests of significance of within-subject symptom changes for NF and sham by study arm. A significant improvement in symptoms was found during NF in both arms of the study (first arm: (F_1,57 = 20.43, p<0.0001; second arm: (F_1,57 = 13.53, p<0.0005). However, reductions during the sham condition were only present among subjects receiving real NF first (F_1,57 =7.44, p=0.0085) compared to no change when receiving sham first (F_1,57 =1.5, p=0.23). This pattern is consistent with a positive carry-over effect driving symptom improvements in the sham condition following NF.

Table 2.

Yale Global Tic Severity Scale Score by treatment, time and study arm

	Mean (95% Cl)		Effect sizeb
	Real-NF	Sham control
Arm 1
Baselinea	25.2 (22.5 to 27.9)	24.8 (19.8 to 29.8)
Endpoint	19.9 (16.5 to 23.3)	23.3 (19.1 to 27.5)
Change	− 5.27 (−7.61 to −2.94)	−1.50 (−3.95 to 0.95)	0.59
Arm 2
Baseline	24.0 (19.8 to 28.2)	20.4 (16.3 to 24.5)
Endpoint	19.5 (15.3 to 23.7)	17.2 (13.2 to 21.2)
Change	−4.50 (−6.95 to −2.05)	−3.18 (−5.52 to −0.85)	0.19

^a:Data are presented as least square means (95% CI) at endpoint and baseline at each study arm and change scores are reported to show within-subject changes in symptoms from baseline to endpoint by condition and study arm.

^b:Effect size is calculated as difference in least square means from baseline to endpoint in Real-NF minus difference in least square means from baseline to endpoint in sham control condition of each study arm, divided by pooled standard deviation at baseline for the respective arm.

The possibility of continued reduction of tics after cross-over for those who received real NF first is also suggested by the number of subjects who were rated as much or very much improved on the CGI-I scale. Specifically, of those receiving real NF first, 5 of 11 subjects (45.5%) were rated as much improved or very much improved compared with 0 of 10 (0%) among those receiving sham first (Fisher’s exact test, p<.05). In contrast, 5 of 10 subjects (50%) receiving NF following sham and 6 of 11 subjects (54%) in the sham group after receiving real NF were rated as much or very much improved, showing no difference in response rates between the groups in the second arm. These analyses suggest carry-over effects are very likely present in this data set. Therefore, an analysis restricted to only the first arm of the study was conducted.

Analyses of first arm of intervention

Analysis of the first arm once again revealed a significant main effect of time (F_1,19 = 17.6, p=0.0005) and a significant treatment by time interaction (F_1,19=5.48, p=0.03). This treatment by time interaction was driven by a significant decrease in the NF condition (F_1,19=22.5 p=0.0001) compared to no decrease among subjects receiving the sham condition (F_1,19=1.65, p=0.21). Specifically, the mean YGTSS Total Tic score decreased from 25.2±4.6 at baseline to 19.9±5.7 at endpoint in the real-NF group and from 24.8±8.1 to 23.3±8.5 in the sham group. The 3.8-point difference (95% CI: −7.15, −0.40) in the YGTSS Total Tic change scores between groups is clinically meaningful and corresponds to a between group effect size of 0.59.

Control task data

There were no significant main effects or interaction terms in the analysis of the control task data. Furthermore, changes in activation of the brain area from baseline to post-intervention were not significantly correlated with symptom changes in the NF group. These data are shown in Figure S6.

Feedback scans

Analysis of control over SMA during feedback runs revealed no main effect of run, a significant main effect of treatment (F_1,18=5.29, p=0.03) and no run-by-treatment interaction. The treatment effect was driven by greater control over the SMA in NF compared to sham (see SI and Figure S3).

Analysis of whole brain activation during feedback runs revealed a significant main effect of treatment after whole brain correction for multiple comparisons, identifying regions more active in NF compared to sham during up- versus down-regulate periods. The most prominent clusters were in striatum and dorsal frontal cortex. See SI, Table SI, and Figure S4.

Debriefing Results

There was no statistically significant difference between treatment groups in terms of whether they believed that they received the experimental intervention, see SI for details.

DISCUSSION

This is the first randomized controlled trial of rt-fMRI-NF in TS. Rt-fMRI-NF is a novel treatment that requires individuals to regulate their brain activity while lying still in the scanner. All subjects in this study completed all NF sessions and outcome assessments demonstrating feasibility of this fMRI intervention in those patients with TS who meet the screening criteria of this study. There was a statistically significant and clinically meaningful reduction in tics during NF relative to changes in the sham control condition. These data suggest that this NF intervention may have clinical potential for the treatment of TS.

Given that neurofeedback is a low-risk, drug-free treatment that would be appealing to many families, further development and testing of the intervention is warranted. Although functional neuroimaging is expensive, two neurofeedback scans require less of a time commitment from families than a course of behavioral therapy for tics (a first-line treatment). It is important to bear in mind that this neurofeedback protocol is highly novel and has yet to be optimized, so future work may increase the efficacy of the intervention. Ultimately, translation of the approach to more broadly accessible modalities may be worthwhile if the intervention proves clinically effective for treating tics in larger trials.

Despite significant reduction of tics after NF relative to a sham control condition, we could not confirm that the treatment worked through the hypothesized mechanism of action. Although we did see greater control over the SMA during NF than during sham feedback in our analysis of the training runs, our registered measure of target engagement, improved control as assessed in the control task runs (our secondary outcome measure) did not differ between the two treatments. Prior neurofeedback studies training patients with movement disorders to control SMA activity have reported mixed findings in terms of improved control over the SMA.(15, 32, 33) In Huntington’s disease, a significant improvement in SMA up-regulation during NF was observed.(32) In Parkinson’s disease, although patients successfully up-regulated the SMA with NF, no significant improvement in up-regulation was observed.(15, 33) Of note, these studies differed from this one not only with regard to the populations studied, they also differed with regard to the training paradigm (only up-regulation was trained), and methods used to assess improved control (assessed during NF not in separate runs). In addition, the functional localization of the target region within the SMA was based on activation during hand clenching in the previous studies, while in the current study patients imitated their tics to allow identification of the part of the SMA controlling tic relevant musculature. Whether the localization approach used in this study captured the most clinically relevant region of the SMA remains an open question.

In any case, given that our secondary outcome measure of control over the SMA did not improve, explanations for the greater symptom improvements during NF than sham that are unrelated to control over the SMA should be considered. One possibility is that the symptom improvements during NF were driven by nonspecific effects such as spontaneous remission and/or placebo effects. Why the sham condition would not share these effects is unclear, but confusion caused by non-contingent feedback may play a role (this could have an effect despite a lack of conscious awareness of the non-contingency).

Alternatively, it is possible that participants did learn to better control their SMA, but our control task scans were not sensitive measures of their control. One factor contributing to noise in our control task measurements could be head motion during these scans. It is a challenge to get clean neuroimaging data from individuals with TS and our efforts may not have been sufficient to provide us with a good measure of control over the SMA. It may be that the participants are able to learn from the neurofeedback despite head motion related noise in the feedback signal because their brain has information about when they are moving (from vestibular, visual, and motor systems) and they can therefore adjust for that noise, while analytic tools for cleaning the data we obtain from control task scans are more limited.

It is also possible that participants were not highly motivated to control the SMA during the control task scans and that we were thus not accurately measuring their ability to control the region. Our finding of greater control during NF compared to sham runs during the training is consistent with this possibility, as the training runs were likely more engaging than the control task scans.

Finally, changes in strategies employed during the control task runs may have added noise to our measurements of control. Consistent with this possibility are the findings from a neurofeedback study that trained SMA activity and compared measures of learned control in subjects who were given motor imagery strategies to a group of subjects who were told nothing about what they were trying to control. Interestingly, while the subjects who were not given strategies demonstrated a significant improvement in their regulation ability, those who were given strategies did not.(34) Thus, better measures of learned control over the SMA may be obtained in studies that blind participants regarding the brain area trained and avoid discussions of potential strategies, allowing subconscious processes to work unimpeded by conscious attempts to control via specific strategies.

A limitation of this study is that it is an early stage trial with a small sample size. Therefore, the results, although promising, cannot be considered evidence of efficacy: larger trials are needed. Future studies with larger samples will also allow for exploration of the possible effects of comorbid diagnoses and medication use. Given the evidence of carry-over effects in these data, and the unknown temporal pattern of clinical changes, a cross-over design is not recommended for future trials. A randomized design including regular symptom assessments for a number of weeks and months following neurofeedback could provide information regarding the temporal pattern of symptom change after the intervention. If tics continue to show reduction for weeks after the intervention as suggested in our recent report, (31) the long-term clinical effects could be substantially larger than the immediate post-treatment effects reported here.

In summary, this first clinical trial of a neurofeedback intervention for adolescents and young adults with TS demonstrates the feasibility and promise of this novel technique, but more work is needed to optimize the parameters of the neurofeedback protocol and to test its efficacy.

KEY RESOURCES TABLE

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Add additional rows as needed for each resource type	Include species and sex when applicable.	Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new.	Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources.	Include any additional information or notes if necessary.
Antibody	N/A	N/A	N/A
Bacterial or Viral Strain	N/A	N/A	N/A
Biological Sample	N/A	N/A	N/A
Cell Line	N/A	N/A	N/A
Chemical Compound or Drug	N/A	N/A	N/A
Commercial Assay Or Kit	N/A	N/A	N/A
Deposited Data; Public Database	N/A	N/A	N/A
Genetic Reagent	N/A	N/A	N/A
Organism/Strain	N/A	N/A	N/A
Peptide, Recombinant Protein	N/A	N/A	N/A
Recombinant DNA	N/A	N/A	N/A
Sequence-Based Reagent	N/A	N/A	N/A
Software; Algorithm	FSL FEAT tool	http://doi.org/10.1006/nimg.2001.0931	RRID:SCR_002823	Used FSL FEAT tool for offline fMRI data analysis.
Software; Algorithm	BioImage Suite 3.5a	PMID: 25364771	RRID:SCR_002986	Used BioImage Suite for offline fMRI data analysis as well as online computation of neurofeedback signal.
Software; Algorithm	MATLAB	MathWorks	RRID:SCR_001622	Used MATLAB to display neurofeedback signal and other fMRI tasks; for offline fMRI data analysis.
Transfected Construct	N/A	N/A	N/A
Other	N/A	N/A	N/A