Jan 25, 2008
(Editor’s Note: Neurofeedback is one of the technologies that people often ask us about. It is a promising intervention in a variety of areas, and has got significant traction in helping kids with ADD/ ADHD. Now, given the significant cost it poses for parents, we need to ask the question: “How Strong is the Research Support for Neurofeedback Treatment of Children with ADHD”? We are honored to present the thoughts of Duke University’s Dr. David Rabiner, a leading authority on the field, on that important issue. As a bonus, you will enjoy his detailed description and suggestions of how to design a high-quality scientific study.)
(Update as of March 2009: Dr. David Rabiner has written an update to the article below based on a newer study. You can read it clicking on link: New Study Supports Neurofeedback Treatment for ADHD)
How Strong is the Research Support for Neurofeedback Treatment of Children with ADHD?
Neither of the two prominent approaches to treating ADHD – medication treatment and behavior therapy – are expected to effect long term changes in the child. Medication treatment induces short-term changes in brain activity that is associated with a reduction in symptoms for many individuals. Behavior therapy attempts to create a set of environmental contingencies that promote desired behavior in the child, but which is unlikely to endure when those contingencies are removed.
In recent years, researchers have begun devoting greater attention to the possibility that children – and adults – may be provided with particular kinds of experiences that may induce alterations in brain functioning that are associated with more enduring changes, i.e., they do not dissipate as soon as treatment ends.
Neurofeedback – also known as EEG Biofeedback – is reflective of this approach and has a history that goes back at least 30 years. It involves providing individuals with real-time feedback on their brainwave activity in order that they may learn to alter their typical EEG pattern to one that is consistent with a focused and attentive state. According to neurofeedback proponents, when this occurs, improved attention and reduced hyperactive/impulsive behavior will result. Furthermore, this change is likely to endure because it reflects a basic alterantion in an important aspect of brain functioning.
In this article, I’ll take a look at some of the controversy surrounding neurofeedback treatment and attempt to highlight some of the strengths and limitations in the existing published research.
– Controversy Surrounding Neurofeedback Research –
Neurofeedback treatment for ADHD has been a source of substantial controversy in the field for many years and remains so today. Although there are a number of published studies in which positive results have been reported, many prominent ADHD researchers feel that given significant limitations to the design and implementation of these studies, neurofeedback should be considered a promising, but unproven treatment.
This position is summarized in the CHADD fact sheet on alternative and complementary interventions, which includes the following statement about neurofeedback:
“It is important to emphasize, however, that although several studies of neurofeedback have yielded promising results, this treatment has not yet been tested in the rigorous manner that is required to make a clear conclusion about its effectiveness for AD/HD. The aforementioned studies can not be considered to have produced persuasive scientific evidence concerning the effectiveness of EEG biofeedback for ADHD.
Controlled randomized trials are required before conclusions can be reached. Until then, buyers should beware of the limitations in the published science. Parents are advised to proceed cautiously as it can be expensive – a typical course of neurofeedback treatment may require 40 or more sessions – and because other AD/HD treatments (i.e., multi-modal treatment) currently enjoy substantially greater research support.”
You can find the complete CHADD fact sheet Here.
What Does an ‘Ideal’ Treatment Study Look Like?
Before reviewing some recent neurofeedback studies, it would be useful to consider what an ‘ideal’ treatment study entails. This will provide a context against which recent neurofeedback studies can be evaluated.
Ideally, treatment studies are designed so that if positive results are obtained, all possible explanations for those results besides the treatment itself have been eliminated. This requires 2 basic elements: random assignment and an appropriate control group.
Imagine that you are testing a new medication treatment for ADHD with 50 children who have been carefully diagnosed. In a random assignment study, whether each child is assigned to the treatment or control condition is determined by chance – you could flip a coin and give the medicine to the ‘heads’ and nothing to the ‘tails’. This insures that any differences that might exist between children who get the medication and those who don’t are purely chance differences. At the end of the study, if those who received the medicine are doing better, you could feel confident that this is probably due to the medicine itself, and not to differences that may have been there before the treatment even started.
What if you didn’t use random assignment, but let each child’s parents choose whether their child is in the treatment or control group? In this case, it is possible that children in the 2 groups differed in important ways before the treatment began. If children who received the medication were doing better at the end of the study, it might be because of differences that were there to start with.
For example, parents who chose the medicine might be more willing to pursue other ways to help their child than those who didn’t. The fact that children who received the medication were doing better at the end of the study might thus have nothing to do with the medicine itself, but reflect other things their parents were doing to help them. No matter how hard you might try to rule out these other possible explanations – and I’m sure you can think of many others – you could never do this with certainty. Thus, I might reasonably doubt that your new medication is really effective.
What about control groups?
Even with the random assignment example above, however, your study would still have an important problem. Because children in the control group received nothing, everyone knows who is being treated and who isn’t. If you ask parents how their child is doing 4 weeks later, this could very possibly bias their reports. Parents whose child received medication may report their child is doing better simply because they expect the medicine to help. Parents of children in the control condition may be biased against seeing improvement because they know their child was not treated.
Thus, if parents of treated children reported more improvement than parents of control children, I could still reasonable question that the new medication was truly helpful. Even if reports from teachers yielded similar findings, I would argue that teachers might have learned when children were getting medication. While this may be a low probability event, you couldn’t completely rule it out. If I was a real skeptic about your new medication, your study would not be all that convincing.
The way around this is to create a situation where no one knows – not the parents, child, teachers, researchers, or anyone else – whether the child is receiving medication. With medication studies, this is relatively simple to do: children in the control condition are given a placebo pill that looks just like the real medication but that has no active ingredients. Because no one knows who is getting medication and who isn’t, possible biases in parents’ and teachers’ reports at the end of the study are thus eliminated. This way, if children in the medication condition are doing better at the end of the study than children getting placebo, it is harder to doubt the benefits of your medication. In fact, there would be no reasonable basis for such doubt, especially if you repeated the study and found the same results.
Three Representative Neurofeedback Studies
With this as a background, let’s take a brief look at 3 recent neurofeedback studies and see how they fare on the critical dimensions of incorporating random assignment and an ‘ideal’ control group.
– Study 1 –
In an interesting study by Vince Monastra and his colleagues [Monastra et al. (2002). The Effects of Stimulant Therapy, EEG Biofeedback and Parenting Style on the primary symptoms of ADHD. Applied Psychophysiology and Biofeedback, 27, 249.] one hundred 6-19 year-olds with ADHD were treated over 12 months. Some parents chose a treatment plan that included medication, behavior therapy, and school consultation. Other parents chose to add neurofeedback to their child’s treatment.
After 12 months, children whose treatment included neurofeedback back were doing better than the other children according to both parents and teachers. They also showed ‘normal’ EEG scans while the other children still had the EEG pattern characteristic of ADHD. Even more impressive was that neurofeedback treated children maintained these benefits after medication was discontinued for a week. You can read a detailed review of this study Here.
– Study 2 –
In a study by Fuchs et al (2003), parents of thirty-four 8-12-year-old children with AD/HD chose either stimulant medication or neurofeedback treatment for their child. After 3 months, children in both groups showed significant and comparable reductions in AD/HD symptoms according to parents and teachers. Laboratory tests of attention also showed equivalent improvement. A comprehensive review of this study is available at Here
– Study 3 –
In an especially interesting study, (Levesque, J., Beauregard, M., & Mensour, B. 2006. Effect of neurofeedback training on the neural substrates of selective attention in children with AD/HD: A functional magnetic resonance imaging study. Neuroscience Letters, 394, 216-221.) twenty 8-12-year-old children with ADHD were randomly assigned to receive neurofeedback treatment – 40 1-hours sessions – or a wait-list control condition.
At the end of the study, treated children were doing significantly better than control children according to parents. They also did better on several objective, laboratory measures of attention. Especially noteworthy was that fMRI scans used to measure brain activity during a complex cognitive task showed significant change for treated children but no change for control children.
You can find a detailed review of this study Here.
So, what’s wrong with these studies?
I suspect the answer to this question is obvious. While the first 2 studies yielded impressive results, neither included random assignment. Thus, as discussed above, it is not possible to conclude that it is the neurofeedback treatment specifically that made the difference. Instead, differences that may have been present before treatment began and/or after treatment started could be responsible. While this may be unlikely, it can’t be entirely discounted.
The third study included random assignment so this is not a problem. It should be noted, however, that with only twenty participants, treatment and control groups are more likely to differ at the start of the study than if a larger sample had been randomly assigned.
The bigger problem is that the control condition is not a very strong one in that everyone knew who received neurofeedback treatment and who did not. This may have biased parents’ ratings, although it is difficult to argue that it could have biased children’s performance on the lab-based attention tests or on the fMRI scan. Perhaps, however, it wasn’t the specific feedback on EEG states that neurofeedback provided, but simply the attention children received during the 40 hours of training, that was the important factor. Although unlikely in my opinion, this can’t be definitely ruled out.
What would an ‘ideal’ neurofeedback study look like?
This may be pretty clear now as well. The most conclusive test of neurofeedback treatment would include random assignment and a control condition that closely matched the neurofeedback condition. For example, children could receive video game coaching from a supportive adult for the same time period. Or, even better, they could do exactly what children getting the neurofeedback were doing but not receive direct feedback on their EEG states. If positive treatment results were still found, it would indicate that obtaining EEG feedback and learning to control one’s EEG state is why neurofeedback treatment works.
I’ve been told by people who know much more about neurofeedback than I that such a study, while difficult to do, is technically possible. I am not aware of such a study having been published; if you are, please let me know.
I should point out that there would be important ethical concerns with such a study. Neurofeedback treatment typically occurs over a period of months. Children who were receiving ‘sham’ neurofeedback would be getting a treatment that no one expected to be helpful for a sustained time period.
A more reasonable alternative might thus be to provide children with a treatment of known efficacy – such as medication – during the neurofeedback trial. If this were given to children in the real and sham conditions it would not create a confound because both groups would be receiving it. After neurofeedback was completed, one could see whether children who got the real treatment were doing better – i.e., did neurofeedback add anything to medication – and whether these benefits persisted after medication was temporarily stopped.
What can we conclude about neurofeedback until such a study is done?
It would be terrific if a study like the one outlined above were completed. Unfortunately, however, I think it is a real possibility, however, that we will never see such a study. This would be a large and expensive undertaking and obtaining funding for it would not be easy.
In the interim, my own view is that existing support for neurofeedback should not be so readily discounted because of the study limitations discussed above. Here’s why I think this is the case.
The way treatment was ‘assigned’ in the first 2 studies summarized is very similar to what happens when parents seek treatment for their child. That is, parents investigate different options and decide which one they want for their child.
This is what happened in these studies – some parents chose neurofeedback for their child and some did not. When neurofeedback was chosen, children were found to benefit on both ‘subjective’ parent reports as well as on more ‘objective’ assessments. This does not mean that neurofeedback would ‘work’ for children randomly assigned to receive it. It also does not mean that reason neurofeedback works by providing specific training in learning how to manage one’s EEG state. It is for these reasons that neurofeedback is understandably regarded as an unproven treatment approach for ADHD at this time by many ADHD researchers.
However, these studies do provide a solid basis for suggesting that if parents choose to pursue neurofeedback for their child, there is a reasonable chance that their child will benefit even though we can’t be sure that it is the specific EEG training that is responsible for the benefits. Thus, although the efficacy of neurofeedback has yet to be conclusively confirmed in a randomized, placebo-controlled trial, it is important to place this limitation in the context of the supportive research evidence that has been accumulated.
Providing this context can help families better understand the strengths and limitations of the existing research on neurofeedback and enable them to make a better informed decision about whether to consider this treatment option for their child.
– Dr. David Rabiner is a child clinical psychologist and Director of Undergraduate Studies in the Department of Psychology and Neuroscience at Duke University. He publishes Attention Research Update, an online newsletter that helps parents, professionals, and educators keep up with the latest research on ADHD, and teaches the online course How to Navigate Conventional and Complementary ADHD Treatments for Healthy Brain Development.
Other articles by Dr. Rabiner: