Attention Defficity Hyperactivity Disorder and fMRI neurofeedback: A randomized controlled trial on the effects of self-modulating anterior cingulate cortex activation levels.

ADHD

Attention-deficit/hyperactivity disorder (ADHD) is the most commonly diagnosed
childhood-onset neuropsychiatric disorder. It is characterized by inattention,
hyperactivity, and impulsivity, either alone or in combination (American
Psychiatric Association, 2000). While 5 to 10% of all school-aged children in
European countries are affected, the disorder may persist into adulthood in one
third of the cases or more (Spencer, Biederman, & Mick, 2007). Individual and
societal costs include impaired academic, occupational, and social functioning,
increased rates of substance abuse, traffic accidents, and persistent
neuropsychological impairments (Biederman, 2004; Secnik, Swensen, & Lage,
2005). Because of the severity and enduring nature of the functional
impairments associated with ADHD, a substantial amount of scientific effort has
been directed on understanding the pathophysiology of ADHD and identifying
effective treatments of ADHD. Both topics will be addressed by this study.

Current treatment of ADHD

While first-line treatment for children with ADHD is the prescription of
psychostimulants (i.e., methylphenidate or dextro-amphetamine), there is no
approved first-line treatment for treating adults with ADHD in the Netherlands.
In general drug treatment in adults has proven to be less effective than drug
treatment in children. A review on the efficacy of medications in ADHD adults
concluded that the response rates to medication were only 50 % when stimulants
were prescribed and as low as 20 % when nonstimulants were taken (Faraone &
Glatt, 2010). Common adverse effects of stimulants include vertigo, decreased
appetite, weight loss, mood lability, tension, and depression (Santosh, Sattar,
& Canagaratnam, 2011). Adverse effects are especially problematic because
treatment is generally long term, as symptoms of ADHD reappear after
discontinuing drug treatment. Also, while medication does improve attention, it
is still unclear if it has a positive effect on academic, occupational and
social functioning in adults with ADHD (Santosh, Sattar, & Canagaratnam, 2011).
Data on the efficacy of alternative treatments as for example cognitive
behavioral treatment in ADHD in adults is still preliminary (Antshel et al.,
2011).

One proposed alternative treatment without adverse side effects is
EEG-neurofeedback. Neurofeedback in general is defined as a procedure during
which a participant learns self-control over some aspect of neuronal
functioning of his brain through getting feedback on it. The aim in general is
to normalize a deviant neuronal pattern, which should also lead to a reduction
of the symptoms of the patient. The goal in ADHD patients is to teach
participants how to control certain EEG signals that are an indicator of
alertness (Sterman, 1996). Recent reviews on EEG-neurofeedback have concluded
that preliminary results are very promising regarding the reduction of ADHD
symptoms and improvement of cognitive deficits (Fox, Tharp, & Fox, 2005;
Heinrich, Gevensleben, & Strehl, 2007; Hirshberg, 2007; Loo & Barkley, 2005;
Rossiter, 2004).
These results have spurred interest into the development of other neurofeedback
methods as well, as for example neurofeedback based on functional magnetic
resonance imaging (fMRI). The advantage of fMRI-neurofeedback over
EEG-neurofeedback may be the higher spatial resolution and full brain coverage
achieved with fMRI, and therefore also a faster treatment response. Functional
magnetic resonance imaging (fMRI) was the first non-invasive imaging method to
provide us with high spatial resolution measurements of blood oxygenation as an
indirect measure of neuronal activity (Bandettini, Birn, & Donahue, 2000), and
has thus advanced our understanding of the human brain considerably over the
last 20 years.

fMRI-neurofeedback

fMRI is a method with a high degree of patient safety, there is no evidence for
hazards associated with increasing exposure (Hawkinson et al., 2011; Schenck,
2000). Since the mid-1990s several research groups have been working on the
development of fMRI real-time techniques, techniques which allow for immediate
data processing and data analysis during fMRI scanning. Current real-time fMRI
procedures include most state-of-the-art data preprocessing and analysis steps
of its classical offline counterpart (Weiskopf, in press; Weiskopf et al.,
2007). Importantly, it has been show that real-time fMRI setups have a safety
level similar to a normal fMRI setup (Hawkinson et al., 2011).
Next many studies have focused on the general feasibility of
fMRI-neurofeedback. As in EEG-neurofeedback the goal is to learn how to
voluntarily modulate some aspect of neuronal activity. Numerous studies have
shown that participants are indeed able to control their brain activation
patterns in very specific ways, and that participants profit from using
fMRI-neurofeedback when learning how to do this (Weiskopf, in press).
Importantly, specific behavioral effects are correlated with specific changes
in brain activation patterns (Weiskopf, in press). Also, first studies with
patients indicate that patients with disorders as diverse as chronic pain,
tinnitus, schizophrenia, psychopathy, parkinson and stroke may experience some
relief from their symptoms after a fMRI-neurofeedback training (deCharms et
al., 2005; Haller, Birbaumer, & Veit, 2010; Ruiz et al., 2011; Sitaram et al.,
2011; Subramanian et al., 2011; Veit, 2009).

fMRI research on ADHD

As the general goal in patient studies is the normalization of the brain
activation patterns which are linked to the behavioral symptoms of this
disorder, one of the most important considerations for setting up an
fMRI-neurofeedback training is which aspect of the brain activation patterns
are most closely linked to the behavioral symptoms. By convergent data from a
variety of sources, including neuroimaging, neuropsychological, neurochemical
and genetic studies, the core symptoms of ADHD have been linked to
abnormalities in the functioning of frontal, cingulated and parietal cortical
brain regions (Bush, 2011). The brain region that has been most consistently
linked to ADHD pathology across all these studies is the dorsal anterior
cingulate cortex (Bush, 2011). Long term structural changes have been shown in
this region (Amico, Stauber, Koutsouleris, & Frodl, 2011; Konrad et al., 2010;
Makris et al., 2007; Seidman et al., 2011; Seidman et al., 2006), and fMRI
research has consistently found a characteristic pattern of hypoactivation
when subjects are performing tasks which are typically challenging to them,
e.g. interference task, continuous performance test, switch task, response
inhibition task (Bush, 2011; Bush et al., 1999; Bush et al., 2008; Cubillo et
al., 2010; Dickstein, Bannon, Castellanos, & Milham, 2006; Schneider et al.,
2010). It has also been shown that this hypoactivation normalizes after
successful treatment with ADHD medication (Bush et al., 2008). Normalization of
this pattern of hypoactivation thus seems to be a crucial aspect in treatment
success.

In the proposed study we want to train ADHD patients how to voluntarily
upregulate the activation level of the dorsal anterior cingulated cortex.
Several previous studies with healthy participants as well as with pain
patients has already shown that in general it is possible to upregulate the
activation level of the anterior cingulate cortex (deCharms et al., 2005;
Hamilton, Glover, Hsu, Johnson, & Gotlib, 2011; Weiskopf et al., 2003).

The principal objective of this study will be to critically evaluate if the
fMRI-neurofeedback training is successful in reducing ADHD symptoms and
improving cognitive functioning. Thus, the first goal is to show that ADHD
patients are able to voluntarily modulate their individual anterior cingulate
cortex activation level. Secondly, it has to be demonstrated that this
modulation has a specific influence on ADHD symptoms and the performance during
cognitive tasks. Finally, it has to be critically evaluated which outcome
measures are positively modulated, and which are the moderating factors for
treatment success. All the cognitive tasks that have been included in the
design are also possible moderating factors.

The second objective of this study is further enhancing the understanding of
the pathophysiology of ADHD. Neurofeedback studies in general are seen as an
excellent tool for investigating the causal influence of abnormal brain
activation levels. As the regional brain activation level is manipulated in a
neurofeedback experiment, the ADHD symptoms and cognitive functioning become
the outcome measure. If one would succeed in influencing behavior through
manipulating brain activation patterns, this would be strong evidence for a
causal role of the abnormal activation levels of the anterior cingulate cortex
in the pathophysiology of ADHD.


Summary objectives

(1) To investigate whether fMRI-neurofeedback training of ACC activation level
reduces ADHD symptoms and improves cognitive functioning.

(2) To investigate if abnormal activation levels of ACC in ADHD patients play a
causal role in ADHD pathophysiology.

This study is a randomized controlled trial (RTC) with blinding of the
participants and blinding of all raters. ADHD adults will be randomly and blind
allocated to one of the following 2 groups:

1. fMRI-neurofeedback training, feedback from dorsal anterior cingulate cortex
(n=10)
2. fMRI-neurofeedback training, feedback from control brain region (n=10)

Each group will receive the same screening, pre- and post-assessment and 6
fMRI-neurofeedback training sessions. The duration of each fMRI-neurofeedback
training session will be approximately 75 minutes. The frequency will be weekly
sessions. The duration of the experiment for each participant from the
selection until the last session will be approximately 8 weeks.
The control group will receive exactly the same as the fMRI-neurofeedback
group, except for the feedback. The control group will receive feedback from a
control brain region, which will be selected such that it is activated by
different tasks than the dorsal anterior cingulate cortex. This selection will
be made based on literature research, previous data, and data from an ongoing
pilot experiment.

In the following part, we will describe detailed the whole procedure from
beginning till end, including the rationale for the proposed measurements.

Location

Selection of subjects, screening for eligibility and assessments, and the
fMRI-neurofeedback training will be performed at the FCDC. All fMRI assessments
will be performed on a Siemens T3 Magnetom MRI scanner.

During the screening session (~ 75 minutes) the inclusion/ exclusion criteria
will be assessed (DSM IV interview, fMRI screening interview, age, IQ Test, use
of medication, significant medical conditions, participation in therapy, and
participation in other clinical trials). Also the following two baseline
measurements will be made:

• A first assessment of the severity of ADHD symptoms the ADHD DSM-IV rating
scale, which is a widely used instrument for determine the severity of ADHD,
fully based on the DSM-IV (American Psychiatric Association, 2000; Kooij et
al., 2005).

• To estimate the intelligence, a shortened version of the WAIS-III-NL
(Uterwijk, 2005) will be administered; the Vocabulary test (~10 minutes) and
the Block Design test (~10 minutes). Validity coefficients for the Vocabulary
and Block Design scores relative to the full form are .88 for Verbal IQ and .83
for Performance IQ (Antshel et al., 2007).

After the screening session randomization will be performed. Randomization will
be used to avoid bias in the assignment of subjects to treatment, to increase
the likelihood that known and unknown factors (expectations of the therapy,
motivation etc.) are evenly balanced across treatment groups and to enhance the
validity of statistical comparisons across treatment groups. Randomization will
be stratified according to ADHD symptom score and IQ score.

An elaborate assessment of the ADHD symptoms will take place during the
pre-training session (~75 minutes). During this session a neuropsychological
assessment and a resting-state fMRI measurement will be completed. The
following neuropsychological tests will be used:

Meta-analytic studies of neuropsychological function in ADHD report moderate to
large effect sizes a sustained attention deficit in ADHD (Willcutt, Doyle,
Nigg, Faraone, & Pennington, 2005). To assess sustained attention we will use
the Sustained Attention dots task (SA-DOTS) as well as the Sustained Attention
to Response Task (SART). Both tasks have been shown to discriminate ADHD
patients from healthy subjects (Marchetta et al., 2008; Slaats-Willemse et al.,
2005; Smilek et al., 2010).

• The SA-DOTS (~15 minutes) is the so-called continuous performance task from
the ANT, a computerized neuropsychological test battery (De Sonneville, 1999).
During the SA-DOTS task the subject is presented random spatial dot patterns
with 3-5 dots. The subject needs to press yes for a 4 dot pattern and no for a
3, or 5 dot pattern. Premature responses, false alarms, misses, the reaction
time and the standard deviation of the reaction time are indicators of the
ability to maintain attention over time. The test-retest reliability of the
SA-DOTS is excellent (0.93 - 0.97, personal correspondence with Leo de
Sonneville), which makes the test suited for measuring differences in the
ability to sustain attention before and after the fMRI-neurofeedback training.
• During the SART (~10 minutes) single digits are presented at a rate of just
over one per second. Participants are told to press a button to every number,
except if that number is 3. As the task is repetitive and apparently easy, it
requires participants to maintain attention.

Another neuropsychological function that has been implicated in the ADHD is
working memory. Differences to healthy participants have been found for verbal
as well a spatial working memory, meta-analytic findings show an effect size of
0.41 to 0.51 for ADHD versus non-ADHD (Willcutt, Doyle, Nigg, Faraone, &
Pennington, 2005). Additionally, working memory capacity has been linked to
successful learning during neurofeedback trainings (Hammer et al., 2012). We
will thus assess verbal working memory with the Digit Span subtest of the
WAIS-III-NL and visuo-spatial working memory using a visuo-spatial Nback task.

• During the Digit Span (~10 minutes) test a sequence of digits is presented
orally. The digit span is then measured by forward- and reverse-order
(backward) recall of the digit sequences. Backward recall is interpreted as a
measure of working memory. In the standard digit span test, the sequences are
presented with increasing length and testing ceases as soon as the participant
makes two consecutive errors. We will use a slightly different procedure, which
was developed recently: all possible trials are presented and a mean score
across all trials is calculated (Woods et al., 2011). Test-retest reliability
for this procedure has been shown to be very good (0.83 for backward
span)(Woods et al., 2011).
• To assess visuo-spatial working memory > NBack task in IMpACT
Finally will assess cognitive interference, as subjects with ADHD usually
underperform on tasks that require cognitive interference (Bush, 2011; Bush et
al., 1999; Bush et al., 2008; Cubillo et al., 2010; Dickstein, Bannon,
Castellanos, & Milham, 2006; Schneider et al., 2010). We will use the
Multi-Source Interference Task (MSIT), a task that has been especially
developed such that it reliably and robustly causes very strong interference
effects (Bush & Shin, 2006). It has also been shown that the MSIT discriminate
ADHD patients from healthy subjects (Bush et al., 2008).

• During the MSIT (~10 minutes) subjects are presented with a visual display of
a set of three numbers (0, 1, 2 or 3). They are asked to report, via button
press, the identity of the number that differs from the other two numbers.
Interference is caused between the value of the target number, the value of the
accompanying numbers, the value of the location of the target number, and the
location of the hand used to answer.

After the neuropsychological testing the pre-training session continues with a
short fMRI scanning session (~20 minutes total), which includes an anatomical
scan (~10 minutes) and a resting-state fMRI scan (~8 minutes). This
resting-state fMRI scan will serve as a baseline measure to estimate the
severity of the ADHD brain pathophysiology. Previous studies have shown that
adults with ADHD show decreased coupling of the anterior cingulate cortex with
other brain regions in comparison with healthy subjects (Castellanos et al.,
2008). Additionally, this fMRI session will have the function of a practice
session, during which subjects can get acquainted with the scanner environment,
and the researchers who will be present during the later fMRI-neurofeedback
training. If subjects are not comfortable in the scanner environment during the
practise session they will be excluded from the study at this point.

• The fMRI session will proceed as follows. Subjects will be carefully
instructed to remove all metal objects before entering the MRI scanner. To
restrict head movements and to limit motion artefacts, the participant*s head
will be fixed by foam cushions and ear clamps positioned behind the neck and
around the head. Participants will also be reminded to keep their head as still
as possible. Headphones customized for MRI experiments will be inserted into
the head coil and will provide isolation from scanner noise. These headphones
will also be used to present instructions to the participants. To accustom
subjects to the scanner noise the anatomical scan will be performed first.
• A high-resolution anatomical MRI scan will be acquired that is optimized for
volumetric measurement of individual brain areas and of gray and white matter
volumes and that can serve as anatomical reference for the functional scans.
Subjects are told to relax and lie still during the anatomical scan.
• Second the resting-state scan will be performed. Resting-state fMRI measures
fluctuations in the Blood-Oxygen-Level-Dependent (BOLD) signal in gray matter
brain areas while the subject is at rest (not performing a task). Participants
will be instructed to relax and remain still with eyes open for 8 minutes in
the fMRI scanner.

During the second week the first fMRI-neurofeedback training session (~75
minutes) will take place. At the beginning of the session subjects are asked to
fill in a short questionnaire on their current state of motivation regarding
the training. Previous neurofeedback studies have shown that state of
motivation is an important predictor of training success (Hammer et al., 2012;
Nijboer, Birbaumer, & Kubler, 2010). Current state of motivation will be
measured using an adapted version of Questionnaire for Current Motivation (QCM)
(Nijboer, Birbaumer, & Kubler, 2010; Rheinberg, Vollmeyer, & Burns, 2001).

• The QCM (~10 minutes) consists of 18 short statements which have to be rated
to which extend they apply on a 7-point Likert-type scale. Four factors of
motivation (mastery confidence, incompetence fear, interest, and challenge) can
be extracted from these 18 items.

After this the fMRI training session (~50 minutes total) will start. Each
training session consists of three parts: a) localizer task (~10 minutes), b)
anatomical scan (~8 minutes), and c) neurofeedback training (~30 minutes). Only
during the sixth and last training session there will be an additional
resting-state fMRI scan (~8 minutes) at the end of the session (see above).

In this study the designated feedback region will be defined individually at
the beginning of each training. At the beginning of the session subjects will
thus be asked to perform the Multi-Source Interference Task (MSIT) as the
localizer task. The individual fMRI data collected during performance of
localizer task will be analyzed immediately using fMRI online analysis software
(TurboBrainVoyager 3.0,
http://www.brainvoyager.com/products/turbobrainvoyager.html). The general
procedure and safety precautions are the same as during the resting-state fMRI
(see above). Visual stimuli will be presented on a screen that the participant
will be able to see by means of a mirror attached to the head coil of the MR
scanner.

The MSIT is chosen as the localizer task because subjects will already be
acquainted with the task, as they have performed it during neuropsychological
testing (see description above). It is also known that this task robustly and
reliably activates the dorsal anterior cingulate cortex, also across sessions
(Bush & Shin, 2006; Bush et al., 2008). Finally, this task has been
successfully used in an ADHD patient group to localize dorsal anterior
cingulate cortex (Bush et al., 2008).

After localizer task and the anatomical scan the fMRI-neurofeedback training
will start (see description in section 5. treatment). The fMRI setup during the
training will be the same as during the localizer task.

After each training subjects will be asked to fill in a short questionnaire
(~10 minutes) on the strategies that they used during this training (Sorger,
2010). Participants will also be encouraged to practise and think about the
mental strategies at home in-between the training sessions. Training sessions
will take place once a week.

One week after the last training session the post-training session (~75
minutes) will take place. All the neuropsychological tests from the
pre-training session will be repeated (SA-DOTS, SART, Digit Span, Nback-task,
MSIT-task, see above), as well as the assessment of ADHD symptoms performed
during the screening (see above).

The fMRI-neurofeedback training will proceed as follows. Before the training
the subjects will be suggested a set of mental strategies that they may use
during the neurofeedback training (same set of strategies in both groups). It
will be stressed that they are always free to choose any strategy that seems to
work, and that they should be guided by the feedback in the selection of their
mental strategy. During the training they will receive visual feedback about
the current activation level (% signal change) of their dorsal anterior
cingulate cortex (experimental group), or the control region (control group).
This feedback will be presented using a visual thermometer display, which will
be continuously updated (every 1.5 second) (Sorger, 2010). The thermometer will
be individually scaled according to the activation level measured during the
previous localizer task (Sorger, 2010). Participants will be instructed to
increase and maintain activation levels to a 50% or to a 100% level, depending
on the cue at the beginning of a 30 second *block*. A similar instruction has
been successfully used in previous studies with healthy participants (Sorger,
2010). Each block of 30 seconds with neurofeedback will be followed by a
resting period of 20 seconds. Subjects will perform 3x8 feedback blocks, with
self-paced breaks after each *run* of 8 blocks. Then, an additional 8 blocks
will follow, during which subjects will be asked to apply whatever strategy
they have learned previously, but now without neurofeedback. They will thus be
asked to transfer what they have learned during the training to a situation
without neurofeedback.

Risks or side-effects are not expected. The burden for the ADHD subjects
consists of an intake, pre- and post-treatment assessments (3 visits of 75
minutes), and 6 visits of approximately 75 minutes for the fMRI-neurofeedback
training. Intake, pre-assessment without fMRI, a treatment-phase and an
evaluation carry the same burden as treatment as usual. The benefit involves of
the a priori chance of positive effect of the fMRI-neurofeedback on ADHD
symptoms.

For all ADHD subjects treatment effect is prospected. The risks of this study
are estimated as very low. Potential benefit of this study is not only expected
for the subjects of this study in terms of treatment response, but for all ADHD
patients in terms of expanding knowledge and extending treatment
opportunities.