No, this article is not about chess. Larry Hirshberg authored a superb defense of EEG Biofeedback (EBF) in application to ADHD that appeared in Expert Reviews of Neurotherapeutics last year but just came to my attention [Expert Rev. Neurotherapeutics 7(4), 315-319 (2007)]. He makes a number of arguments that we should all have in our grab bag when challenged on neurofeedback.

First of all, the basic question of whether operant conditioning on dynamic aspects of brain function is even possible is now supported by evidence from the neighboring field of brain-computer interfaces (BCI), which is also dependent on a learning paradigm. Additional evidence is furnished by real-time functional MRI feedback (rtfMRI). “While this research has generated considerable scientific and popular interest and excitement, it is less often remarked that it validates the fundamental premise of EBF…”

The analogy is particularly close when the BCI technology employs EEG signals derived at the scalp. An initial period of feedback-based learning is required for the subject to gain control of the cursor. “This feedback-guided learning is EBF-training, pure and simple.”

Countering the argument that neurofeedback needs to establish itself against the prevailing standard approach of stimulant medication, Hirshberg retorts that “a large percentage of patients with ADHD either do not receive or do not sustain medical treatment.” So EEG biofeedback is at minimum an evidence-based treatment that can be “employed in practice when medical treatment fails or is not accepted by the patient.” Studies indicate that some 50-87% of children diagnosed with ADHD in the US “either do not begin, or fail to continue medical treatment.” It has been estimated that the average length of time that stimulants are consumed by youngsters is less than six months, so even under these limiting assumptions the potential for neurofeedback is huge.

Hirshberg also challenges the view that the scientific evidence supporting EBF for ADHD is still marginal. “Overall, more than 20 studies have been conducted involving over 700 subjects. Nine have been controlled trials, involving over 400 subjects, with comparison to sham treatment, nontreatment and stimulant medication controls; these include five randomized controlled trials.” This vastly exceeds the evidence that is available to support many medical treatments, such as, for example, the use of anti-depressant medication in children. And, as Hirshberg points out, it exceeds the evidence for many off-label uses of medications and for the common practice of “medication stacking” in application to ADHD.

Unfortunately, two of the most recent studies that approximate the ideal in terms of research design are referenced only as conference presentations rather than as publications. By that standard, Hirshberg could also have cited the Cal Poly study that we sponsored back in 1994-5, which involved a three-group design with a neurofeedback cohort, a wait-list cohort, and a cognitive training cohort. This study was reported at the 1995 Conference of the American Psychological Association. Significantly, we worked with an already medicated population, so the study provided evidence for the additive value of neurofeedback for a population already managed to community standards. Both clinically and statistically significant improvement was observed on the hyperactivity scale of the Conners Parent Rating Scale (CPRS) even in these already-medicated children.

In addition to the controlled studies that are coming along, there have been a number of corroborative observational studies over the years. Hirshberg points out that “recent meta-analyses comparing the results of observational studies versus randomized controlled trials…reveal that results from the two approaches to research are generally concordant.” Referring to Benson and Hartz, who reviewed 136 reports related to some 19 medical treatments, these authors concluded that in only two out of the 19 cases did the combined effect size found in the observational studies lie outside of the 95% confidence interval of the randomized controlled trials (RCTs).

Consequently, Hirshberg suggests, “a balanced approach to the status of scientific evidence should take into account results from observational as well as controlled trials…” The bias of accepting only the evidence from RCTs “appears itself to represent an opinion unsupported by the evidence base.” One can push this argument even further. No one, for example, raises the issue of placebo response with regard to the BCI research or even the fMRI feedback study. In the case of the BCI research one is a witness to a progressive learning curve. In the case of the fMRI feedback study of chronic pain, the results are repeatable from session to session.

What has perhaps gone relatively unnoticed in the neurofeedback field is that we have gradually progressed to the point where behavioral change is routinely observable in a single session. That is to say, the learning curve is on view in real time, just as in the case of BCI. Moreover, these results are obtainable with children who are no more wise to the program than were Barry Sterman’s cats. We are talking about young, severely impaired autistic children. Anyone who does this work well is literally immersed in tiny, corroborative bits of evidence that support our overall claims for neurofeedback–the learning curve, first of all, but also the contingency on reinforcement parameters, etc., and finally the myriad small indicies of progressive behavioral change that come with enhanced capacity for self-regulation. Merely witnessing this process affords compelling evidence for the ultimate claims.

So on the one hand the critics goad neurofeedback professionals into doing ever more careful research, while they themselves remain entirely absent from the stage. One must ask, why are these people not even curious enough to witness the process? Why do they not ask to experience this adventure on their own heads? The lack of curiosity is the worst indictment one can make of a presumptive scientist, short of scientific fraud. (OK, there may be some of that to be found, too, among our critics, in their misrepresentation of the facts.)

Hirshberg calls for a more nuanced perspective on neurofeedback than the ‘supported-versus-unsupported’ dichotomy. There must be that middle ground in science where techniques can migrate toward more solid grounding. Instead the ADHD brain trust in the US has been solidly hostile to any formal inquiry into EEG feedback over the years. The only exception is a recent one, the study to be undertaken by Eugene Arnold. In a review of investigational remedies for ADHD Arnold called for formal studies of EEG neurofeedback some years ago, but that never went anywhere. (We had a long conversation about EEG feedback at the time.)

And on the clinical side, Hirshberg reports that the largest group of ADHD clients seeking neurofeedback consists of those who tried medical treatment with equivocal results, and the second group consists of those who refuse such medical treatment categorically. What then, is one to do? If the first-line treatment fails, one resorts to other alternatives that may not be as well supported. And similarly if the first-line treatment is refused, the therapist enlarges the scope of options for the client. Hirshberg argues compellingly that “an informed approach to evidence-based practice with ADHD would fully and wholeheartedly recommend such a role for EBF in the treatment of ADHD.”

Additionally there is the prospect that with EEG feedback the trainees may be able to significantly reduce their medication requirement. And benefits are often reported for aspects of behavior that are not addressed by stimulant medication: “…we often see significantly improved emotional regulation and lability and reduced anxiety.”

Having made this compelling case for neurofeedback for ADHD, Hirshberg nevertheless prefers to err on the side of conservatism when he concludes that “EBF is best viewed as an alternative approach to addressing the primary symptoms of the disorder that has substantial, but not yet conclusive, evidence of efficacy.” Even more conservatively, he insists that there is only “preliminary evidence that any gains that may be obtained will last.” Oh, really? Isn’t this the point of the whole enterprise? If we weren’t persuaded that the gains hold, then the game would not be worth the candle. The evidence for neurofeedback is only as strong as the evidence that the gains are retained over the long term.

This evidence has in fact been in hand since the very beginning. When Sterman discovered the heightened tolerance of his trained cats to the toxin hydrazine, the cats had not seen a neurofeedback training session in months. Lubar confirmed that his early ADHD subjects retained their gains for ten to fifteen years. In our publication on neurofeedback for mildly mentally retarded twins, the improvements were tracked for five years and shown to hold. The first person to tell us that we had “cured” her migraines with the neurofeedback had not had a migraine for three years. In our addictions treatment study we tracked the results for three years and found them to hold. Peniston tracked his results for ten years. We can also go back to the neurofeedback analog, the BCI technologies. Once a chimpanzee learns the skill of cursor movement, the skill is largely retained. The same holds for most rehabilitation modalities.

Hirshberg makes the case for including observational studies along with controlled studies in the evidence base to be considered. This leaves out the largest resource of all, which is actual clinical experience, which extends over more than three decades and by now involves thousands of clinicians worldwide. As both a practitioner and a researcher, Hirshberg has by this time surely resolved any ambiguity on the matter of neurofeedback efficacy in his own mind. Why is this not ok to say? The cumulative wisdom and conviction of a skilled clinician surely matters.

A person who observes the learning curve in a BCI experiment ends up thoroughly convinced that a learning process has been witnessed. Why should our clinical evidence be given less weight than that, particularly when the same report comes from such a variety of clinical talents and involves such a variety of feedback (and stimulation) techniques. These clinicians, who may agree on virtually nothing else, will all support the proposition that through neurofeedback they are able to be more clinically effective than ever before. Why shouldn’t this just be said outright?

Finally, I want to take issue with the continued characterization of neurofeedback as an alternative treatment for ADHD. What is missing from Hirshberg’s treatment is any consideration of a model of ADHD that might make neurofeedback efficacy an expectation. It should be clear by now that the core of ADHD is to be found in the temporal organization of our neuronal networks, particularly as regards executive function, the modulation of arousal, the maintenance of vigilance, and emotional regulation. If there is a neuromodulator deficit in ADHD, it certainly has been elusive.

On the other hand, if the heart of the problem lies in network functional organization, then neurofeedback should be considered the primary remedy, and pharmacology the secondary alternative. The fact that the vast majority of ADHD children no longer derive substantial additive benefit from stimulants after neurofeedback training shows that drugs are largely just a patch on the real problem. Once the ‘real’ remedy is at hand, the medications largely become superfluous. This would not happen if the ADHD child were truly handicapped by an intrinsic neuromodulator deficit.

So, by the argument of medication displacement the pharmacological model is dethroned, and neurofeedback must be given priority in our model considerations. The reduction in medication requirements post-feedback is also strong evidence for the permanence of the training effect. Hirshberg came so close to nailing the coffin shut on the neurofeedback deniers. Instead his potentially powerful message dissipates in a mere whimper because he ultimately lapsed back into dichotomous thinking when he grades the cumulative evidence as “less than conclusive.” By defaulting back to dichotomous thinking, Hirshberg is buying into our critics’ mindset and allowing them to control the terms of debate. The critics would like nothing more than to set a high bar, with respect to which we would always fall somewhat short.

Dichotomous thinking can be avoided if we consider neurofeedback in terms of a more nuanced hierarchy of claims. There is first of all the claim that neurofeedback techniques are able to alter behavioral states in the moment. Of this there is no longer any doubt at all. One can even point to all the evidence from alpha-theta training in support of this proposition. Then there is the claim that learning occurs in this process, and hence there may be a residual effect beyond the actual session. On this point there is convincing proof as well. Then the claim is advanced that EBF can be effective with respect to the cardinal symptoms of ADHD. The existence proof here is already iron-clad. Additionally, the evidence that neurofeedback can impinge on hyperactivity, impulsivity, inattention, and distractibility accumulates every day in our offices, as documented with CPT tests, other cognitive skills tests, and behavioral assessments.

Beyond this, the questions lie in the analog domain. The issues are no longer black and white. The question ‘How effective is neurofeedback in comparison with the alternatives?’ cannot be answered simply. These are multi-factorial issues, and matters are also in continuing flux because of ongoing developments in the field.

The debate must also be moved onto ground that is more favorable to our cause. The way to do that is to assert the primacy of the network model in understanding ADHD. As Einstein said, it is the theory that tells us what we may believe. As soon as we talk in terms of the network model, we control the terms of debate. This is the frontier of neuroscience, and with neurofeedback we have not only a powerful remedy but also a sensitive probe into the functioning of our neural networks. We have the very tool that neuroscience has badly lacked–a way of subtly influencing the states of the system without upsetting them.

Siegfried Othmer, Ph.D. 

This entry was posted on Thursday, June 12th, 2008 at 9:15 pm and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.