Much proposed research on neurofeedback has faltered over the years on the issue of uniformity of approach. The protocol to be investigated needs to be narrowly constrained or the research will be criticized for a lack of specificity. We have had a number conversations over the years with researchers who were willing to give Neurofeedback research a go, provided we would give them a fixed protocol to work with. In recent years, we have been increasingly unwilling to do this, and by now the point has been reached where such fixed protocols are plainly inappropriate.
The best neurofeedback requires one to react to how the client is reacting to the training, and to make appropriate adjustments. The moment we know this, it becomes ethically questionable to proceed in a manner that sweeps such particularity under the carpet. As it happens, however, the way we proceed in practice is not very different from the way a psychiatrist might proceed in optimizing medication. The choice of medication is often driven more by side effects than by bare-bones efficacy, particularly in the case of the most common medications, the antidepressants. And in many instances medications are combined in various ways. Research by classical methods is not of help when it comes to polypharmacy, and it is not of help with regard to handling side effects. To some degree, therefore, ‘Personalized Medicine’ has already emerged in pharmacotherapy, but this has not yet succeeded in feeding back to rewrite the rules for research. A recent analysis of the status of the science at the FDA, requested by FDA Commissioner Andrew von Eschenbach, came up with the recommendation that the FDA could help to define personalized medicine. “This is the science that the FDA can really take a lead on.” [Science, 318, p.1537, 7 Dec 2007] In neurofeedback, we are in the same boat.
The equivalent of a medication side effect in neurofeedback is the non-optimal response, obtained most readily by training at the wrong reward frequency for the particular person. The client may not find relief for his migraines, and may even find that training at the wrong reward frequency can trigger a migraine. Research on any clinical method that requires individualization of the protocol must allow sufficient latitude to the clinician. The science, then, cannot be totally prescriptive. Rather, it must become more observational, or descriptive. In fact, the science of neurofeedback would be greatly served if someone schooled in research methods would simply observe the clinician at work, bringing their scientific rigor into the clinic. The researcher would make no attempt to specify what is to happen, but rather would simply characterize and quantify the outcomes. The client/therapist dyad is not to be interfered with in any way by the researcher. It is this dyad that effectively becomes the object of study. Research on the efficacy of certain surgical procedures already follows this model.
This approach to neurofeedback research offers a badly needed corrective to the view that neurofeedback is reducible to a procedure that can be simply “manualized” and replicated at will by anyone capable of reading the manual. A similar over-simplification is at work on the diagnosis side, where diagnostic specificity ends up burying a lot of clinical complexity. An apparent order is imposed upon nature that does not in fact exist. We already have an example of personalized medicine in QEEG-based Neurofeedback training. The protocol is determined from observed EEG anomalies. These are rarely found in isolation, so a hierarchy of approach is called for. Such hierarchies are typically established empirically at the clinician level, but any success in any such approach is still deemed to contribute to the proposition that QEEG-based protocol decision-making is helpful. The individualization of training approaches here has never been seen as a problem for research.
Now when it comes to protocol-driven training, adapting the protocol to the client is somehow seen as more problematic. In particular, it is problematic because client/clinician interaction is involved in driving the protocol refinement. Clinician skill is an issue because it is not quantifiable, nor is it replicable necessarily in other settings. Research conducted top-down in the old-fashioned way would like to see such variables taken out of the equation. Observational science, however, would first of all concern itself simply with outcomes. Once neurofeedback is established on the basis of outcomes research, the agenda can expand into a closer examination of clinical methods.
What is clearly upside down right now is that with standard research methodology, it is the researcher who will tell the technician involved in the study what to do and how to do it. This has historically never worked very well in feedback research, and the history of such failures is extensive indeed. Those who are prepared to delve into Neurofeedback research now are surely oblivious of this history. It is therefore most unlikely that the researcher will draw the conclusion that his methodology is unsuitable to this purpose. There is one area in medical research that must rely on observational science, and that is adverse events reporting. Of interest is the rate at which certain adverse outcomes appear in the target population. Right now, it is said that the FDA knows neither the numerator (the number of adverse events) nor the denominator (the number of people treated with a particular medication). But the remedy is clear: more observational studies are needed, and the methodology is straight-forward. Questions of mechanisms of adverse outcomes can all be left to later, when we know how serious the issue is.
A similar formalism can be applied to the practice of neurofeedback. Let the researcher simply track systematically what happens in neurofeedback offices. Any subsidiary research objective can be left for later, when it is more apparent what priority it should have. It is just possible that the movement toward personalized medicine will prove to be the end of the road for standard research methods, involving group studies based on assumed homogeneity, and imposing uniform procedures under controlled, blinded conditions with random assignment to groups. Long live ‘Personalized Medicine.’