Comparing tomographic EEG neurofeedback and EMG biofeedback in children with attention-deficit/hyperactivity disorder

Highlights

• Tomographic neurofeedback training (NF) of ACC activity is compared to EMG-biofeedback.

• Blinded ratings of ADHD symptoms indicate improvements after both treatments.

• Larger improvements on visuo-motor and coordination tasks after EMG-biofeedback.

• Learning regulation across sessions is found for EMG-biofeedback but not for NF.

• Only NF shows a trend toward normalization of the trained frequency bands.

Abstract

Two types of biofeedback (BF), tomographic electroencephalogram (EEG) neurofeedback (NF) and electromyographic biofeedback (EMG-BF), both with phasic and tonic protocols, were compared for treatment effects and specificity in attention-deficit/hyperactivity disorder (ADHD). Thirteen children with ADHD trained their brain activity in the anterior cingulate cortex (ACC), and twelve trained activity of arm muscles involved in fine motor skills. In each training session, resting state 24-channel EEG and training performances were recorded. Both groups showed similar behavioral improvements and artifact reduction in selected conditions, with no significant advantages despite medium effect sizes on primary outcomes for NF. Only the EMG-BF group, however, showed clear improvement in training regulation performance, and specific motor coordination effects. The NF group tended to present individual normalization of trained frequency bands in the ACC during rest across training. The results provide evidence for some specific effects in our small sample, albeit only to a small extent.

Introduction

Attention-deficit/hyperactivity disorder (ADHD), with a worldwide prevalence of approximately 5.2%, is one of the most frequent disorders in psychiatry (Polanczyk et al., 2007, Steinhausen et al., 1998). The core symptoms of ADHD are inappropriate levels of inattention, impulsiveness, and hyperactivity (Barkley, 1997). In addition, children with ADHD often have comorbid motor coordination problems (Fliers et al., 2008, Kadesjo and Gillberg, 2001, Rommelse et al., 2007, Slaats-Willemse et al., 2005, Steger et al., 2001).

With regard to the core symptoms of ADHD, several treatments are typically used. Although the use of stimulant medication is widespread, only about 70% (Barkley, DuPaul, & McMurray, 1991) of children with ADHD respond to pharmacological treatment. In addition, side effects, reluctance to take medication, and the lack of clear positive long-term effects are serious limitations of this treatment (Banaschewski et al., 2006). Consequently, there is a strong demand for alternative behavioral treatments such as neurofeedback (NF), which, based on learning of regulation or operant conditioning of brain activity, is considered an alternative or additional treatment (Heinrich, Gevensleben, & Strehl, 2007). NF is geared toward building the self-control of neurophysiological functions which are altered in ADHD (Doehnert et al., 2013, Monastra et al., 2001) and to normalize them, but may also support compensatory regulation strategies (Gevensleben, Rothenberger, Moll, & Heinrich, 2012). The regulation of theta (4–8 Hz) and beta (13–20 Hz) frequencies as well as the training of slow cortical potentials (SCP) are typical NF training protocols used for the treatment of children with ADHD (Drechsler et al., 2007, Gevensleben et al., 2009a, Heinrich et al., 2004, Leins et al., 2007, Lubar et al., 1995, Strehl et al., 2006a, Thompson and Thompson, 1998), which have been used in an adapted and tomographic variant in the study presented here (see also Liechti et al., 2012).

There is increasing evidence that training the self-regulation of neurophysiological parameters through NF, using scalp electroencephalogram (EEG) from a single channel (conventional NF), improves ADHD symptoms (Doehnert et al., 2008, Drechsler et al., 2007, Gevensleben et al., 2009a, Heinrich et al., 2004, Kropotov et al., 2005, Leins et al., 2007, Strehl et al., 2006b). Correspondingly, a meta-analysis (Arns, de Ridder, Strehl, Breteler, & Coenen, 2009) reporting large effect sizes for inattention and impulsivity and a medium effect size for hyperactivity, even when compared to control groups, recommended NF as an “efficacious and specific” ADHD treatment. In contrast, a review by Lofthouse, Arnold, Hersch, Hurt, and DeBeus (2012) considered it only as “probably efficacious”, and a recent meta-analysis (Sonuga-Barke et al., 2013) reported only a trend for probably blinded ratings (mostly teacher ratings). As discussed in our previous publication (Liechti et al., 2012), a more efficient approach than conventional NF might be the training of intracerebral activity in specific brain regions affected in ADHD such as the anterior cingulate cortex (ACC). This brain region has consistently been implicated both in EEG-based (Albrecht et al., 2010, Fallgatter et al., 2004) and other imaging studies (metaanalyses, Cortese et al., 2012).

An important aim in NF research is to demonstrate and understand the specific mechanisms of action of training protocols and their impact on the training outcomes, and to clarify the nature of unspecific mechanisms. One approach to investigate specificity, which was pursued in our previous paper (Liechti et al., 2012), is to examine the relation of the individual learning of neurophysiological regulation to the training outcome. Another approach is to compare NF effects to those of a control training or group, which is pursued in this paper comparing the same NF group with an electromyographic biofeedback (EMG-BF) control group. An active control condition, which consists of a comparable amount and intensity of cognitive demands and patient–therapist interaction, allows to disentangle the specific and unspecific effects of NF treatment. This approach controls for unspecific effects induced by the NF setting, such as patient–therapist interaction, immediate feedback, reward, systematic training to sit still, attentional aspects of the training, expectations generated by applying electrodes, and being connected to a computer (Arns et al., 2009, Brandeis, 2011). This is contrasted with a waiting list group, which is a passive control condition eliminating only those unspecific confounds due to elapsed time and test repetition.

Some studies also reported protocol-specific neurophysiological changes (Brandeis, 2011), particularly for SCP training (Doehnert et al., 2008, Heinrich et al., 2004, Wangler et al., 2011). Some specific results have also been reported for frequency band training protocols. For example, Gevensleben, Holl, Albrecht, Vogel, et al. (2009) found that behavioral outcome after theta/beta training correlated with theta decrease.

Only few studies have examined associations between the training regulation performance during the training and behavioral improvement, but most of them provided at least some evidence for significant relations (most recently, Gevensleben et al., 2013; for reviews, see Drechsler, 2011, Moriyama et al., 2012). In our previous publication (Liechti et al., 2012), improved clinical ADHD symptoms and differential ACC modulation were also found after NF, but there was no or little training regulation success and consequently no relationship between the training success and the training outcome. For these reasons, we concluded that unspecific or secondary NF effects such as artifact control account for much of the clinical improvement, but in order to clarify the remaining specific treatment effects, a comparison with an active control group is essential.

Control conditions are critical to determine specific effects of NF in randomized controlled trials. The choice of the appropriate control condition for NF remains a matter of debate (Gevensleben et al., 2012, Lofthouse et al., 2012b, Loo and Makeig, 2012). A sham NF group with placebo feedback utilizing the same setting and interface for training represents the most powerful control group in some respects (Lansbergen et al., 2011, Logemann et al., 2010). Differences in clinical outcome between regular and sham NF training which are equivalent in all other aspects of the setting, can be attributed to the specific effects of learned regulation of the targeted brain activity. However, sham NF training may induce higher drop-out rates (Arns et al., 2009) and reduce the active effort. In addition, implementing a sham condition in clinical research with ADHD children is critical from an ethical point of view. Another biofeedback (BF) method with genuine feedback and a possible therapeutic benefit, such as the feedback of motor activity, seems to be a preferable alternative. In addition, blinded ratings (Sonuga-Barke et al., 2013) plus evaluation of blinding (Gevensleben, Holl, Albrecht, Vogel, et al., 2009) may offer reasonable control for unspecific expectancies. EMG-BF has been used in several studies to improve muscle relaxation and reduce hyperactivity in children with ADHD (for reviews see Arnold et al., 2011, Cobb and Evans, 1981, Lee, 1991). In the present study, EMG-BF focuses directly on the improvement of fine motor skills and motor regulation, which is often impaired in children with ADHD (Pitcher et al., 2002, Pitcher et al., 2003) and may therefore represent a meaningful treatment for this group. In addition, this EMG-BF training provides well-matched control conditions for NF training (for a more detailed description of the method, please refer to the case description of this control condition by Maurizio, Liechti, Brandeis, Jäncke, and Drechsler (2013), and to the supplementary material).

So far, only one study has used EMG-BF as a control condition for NF (Bakhshayesh, Hansch, Wyschkon, Rezai, & Esser, 2011). In this study, the participants had to reduce the EMG amplitude of the feedback signals of the forehead musculature. Significant improvement of ADHD symptoms was reported after both training conditions; there was more improvement after NF for inattention symptoms on parent rating scales and reaction times in neuropsychological tests, but no significant differences were found on teacher ratings or other measures, suggesting that NF may only have limited specific behavioral effects when unspecific factors are controlled for. The specific mediators of response of NF and EMG-BF, however, are still unknown and need to be investigated further. For this reason, in the present study, we compare NF and EMG-BF in two groups of children with ADHD and investigate training regulation performance, artifact control, and resting EEG changes in the course of the training and behavioral and neuropsychological changes between pre- and post-assessments.

We hypothesized that ADHD behavioral symptoms would be mitigated in both groups, although we expected larger effects for the NF. Through this new training technique, we had expected that children would learn to specifically regulate their EEG in a brain region most affected in ADHD and thereby show stronger improvement on behavioral rating scales and neuropsychological tests tapping into attention compared to the EMG-BF treatment. However, due to unspecific effects, some ameliorations in the EMG-BF group were also expected, in accordance with Bakhshayesh et al. (2011). Following our previous results providing more evidence for regional normalization at rest than for learning of regulation (Liechti et al., 2012), we expected the individual normalization of resting EEG activity in the ACC to be specific to the tomographic NF treatment. For the EMG-BF group, changes in the resting EEG and a stronger improvement on neuropsychological tasks related to fine motor skills and bimanual coordination were expected.