Fifty ways to leave your love… of the Placebo Model of Neurofeedback

by Siegfried Othmer | December 28th, 2017

Neurofeedback has now existed for more than half a century, and it is flourishing in the clinical arena. Many academic researchers, meanwhile, are still stuck at first base worrying about the placebo as a complete explanation of neurofeedback efficacy–the proposition that neurofeedback is nothing but an expensive placebo. Looking at this collectively, it appears that the the placebo hypothesis is drawing much more interest among some academics than what is actually happening in the clinical realm. To those of us inhabiting the real world of clinical practice, this is such a ludicrous spectacle that it is difficult to take seriously. Researchers tend to orient to the published literature and thus listen only to each other. That misses most of the ferment, vitality, originality, and diversity of the field, which happens to reside almost exclusively in the world at large rather than in the formal environs of the research laboratory. Much of the resulting novelty and diversity has not found its way into the journals for the simple reason that clinicians by and large don’t publish their work. They network with other clinicians at conferences and through the Internet.

Clinicians are not oblivious to the problem of the placebo in principle. But as a practical matter, the placebo hypothesis no longer carries weight with them in connection with neurofeedback. In their minds, that issue has been put to bed. That is for an untold number of reasons that emerge directly from clinical practice, as enumerated below.

The issue is this: For neurofeedback to be reducible to the placebo, it has to be able to account for everything. It cannot just explain some things. If evidence can be adduced that cannot be explained by way of the placebo hypothesis, then the hypothesis expires. We therefore argue on the basis of the “Black Swan” principle: “In contradiction of the proposition that all swans are white, one merely has to observe a single black swan.” By means such as these, the placebo hypothesis can be extinguished in a variety of ways without the benefit of any additional controlled studies having the explicit objective of testing the placebo hypothesis.

Of course that has not always been the case. Placebo-controlled studies did have be done at the outset, just to launch the field in the first place. But once such experiments are successful, there is no need to keep doing them. One black swan is enough to allow clinicians to proceed ethically. Once that happens, of course, one encounters black swans all the time… Beyond that, in the real world we don’t live by absolutes most of the time. Whereas the researcher is after rigorous, definitive judgments, the clinician navigates by means of best judgments. For those of us who have been involved with the clinical practice over the decades, the evidence has always been overwhelmingly in favor of the proposition that neurofeedback is clinically effective not only in the statistical but also the practical sense. So the reader may also view the following items in terms of the “likelihood of validity” as opposed to the categorical contradiction of the placebo hypothesis. After all, the placebo effect does exist, and it plays a role in all clinical work. In the neurofeedback perspective, the placebo is a compounding factor, not a confounding one.

Cumulatively, the evidence from the clinical domain can reach any arbitrary level of certainty, which realistically cannot be matched by any number of independent group studies. It will be seen in the following that the clinician has plenty of ways to rule out the placebo hypothesis without running formal studies. We begin, however, with the formal studies that kicked everything off in the first place.

Fifty ways to annihilate the placebo hypothesis of neurofeedback efficacy:

  1. Sterman’s original cat research, which first established neurofeedback, was a totally blinded, placebo-controlled, balanced design–by pure happenstance! This work established an existence proof for EEG biofeedback that violates the placebo hypothesis. It does so for all time—at least for cats.
  2. The cat experiments were successfully replicated with monkeys, who likewise demonstrated successful neurofeedback that could not have involved a placebo element. That’s cats and monkeys. What about people?
  3. Successful neurofeedback with infants does not fit the placebo model. Here’s a mother’s report on her three-week old daughter: “This was a very colicky baby, with severe constipation and poor sleep, who would cry inconsolably for hours and hours in a row. She calmed down in her very first session, and stopped crying eighteen minutes into it. Sleep improved and constipation resolved.” There is no placebo explanation for this.
  4. Successful neurofeedback with a person in vegetative state (relief from constipation; regularization of the menstrual cycle) cannot be explained by the placebo hypothesis.
  5. Recovery from Level 2 coma: Surveying clinical experience over a number of years, Margaret Ayers reported that out of 32 people in Level 2 coma for periods greater than two months, 25 could be recovered within one or two sessions; two more could be recovered within several additional sessions. The remaining five could not be brought out within six sessions. This finding categorically disposes of the placebo argument.
  6. Reversibility of the protocol implies instrumental agency outside of the person. Lubar published several controlled studies on ADHD and epilepsy that utilized reversal designs. Instead of down-training of the theta-band, up-training of theta amplitudes was instituted in a second phase of the training. Here the expected reversibility of the training effects was demonstrated for both conditions, as symptoms once again got worse. (Reversibility wasn’t necessarily to be expected, as learning occurs in this process.)
  7. EEG biofeedback doesn’t just have effects on brain-based dysfunctions. It also evokes discernible state shifts in real time within the training session. These state shifts exhibit obvious dependence on the training frequency, to which the client is blind. The parametric dependence rules out the placebo model with respect to these state shifts.
  8. Reversibility in real time: Toggling back and forth between two target frequencies can reversibly shift state. For example, one frequency may produce the sensation of hunger while another produces the sensation of satiety. This phenomenon is repeatable within a particular individual. If challenge A and challenge B have systematically different effects, they cannot both be assigned to placebo.
  9. Reversibility also holds with respect to clinical issues: Training at one frequency may, for example, induce a migraine aura, whereas reinforcement at a neighboring frequency will quickly quench the aura. This is a reversible phenomenon. Since the same “good” frequency that quenches the aura also facilitates the client becoming migraine-free over the course of a few training sessions, the placebo model is contradicted not only for the state shift but for the remediation as well. Both argue for the involvement of instrumental agency.
  10. Reversibility in a command performance: A dramatic case of reversibility of hand tremor was demonstrated by Lisa Tataryn to Scott Makeig at his neurosciences laboratory in La Jolla many years ago. Tataryn calmed the man’s hand tremor with one protocol, restored the hand tremor with another, and calmed it again with the original protocol. All this transpired within thirty minutes. Extended training is expected to calm hand tremor over the longer term.
  11. A similar case of reversibility was observed by Sue Othmer as she was working with a person with Dissociative Identity Disorder. As the target frequency was walked down the client regressed to ever younger ages, eventually speaking in the high voice of a child. Walking the frequency back up reversed the entire process.
  12. Parametric dependence of protocols: Clinical success is often contingent on adaptation of the target frequency to the client. This is observed particularly for conditions in which the brain is unstable (e.g., epilepsy, migraine, asthma, bipolar mood swings, vertigo, and trigeminal neuralgia).
  13. By the same token, clinical success for conditions of brain stability is strongly correlated with electrode placement.
  14. Neurofeedback may induce adverse physiological responses, in particular those related to the client’s particular vulnerabilities—agitation, anger, irritability, migraines, panic, or rage. Clearly these consequences were not intended by the client, and hence cannot be assigned to the placebo.
  15. Successful neurofeedback with profound and thorough-going skeptics cannot be explained in terms of the placebo model. Ironically, sometimes these skeptics appear almost to prefer being validated in their skepticism than to benefit from the training. In any event, they come to the experience well-armored. Often it is an unexpected initial adverse reaction to the training that persuades the critic of the validity of the procedure, precisely because it is easy to assign an initial positive change to the placebo.
  16. A fixed mathematical relationship has been identified for the optimum frequencies at different placements. These relationships are universally observed, and yet the trainee is always blinded to training condition with respect to target frequency. The very existence of these fixed relationships across populations serves to discredit the placebo hypothesis.
  17. Often remediation of a particular symptom is achieved only with the second or third protocol downstream. In these cases, the early protocols can be analogized to a placebo washout-period. The second or third trial cannot then also be assigned to the placebo. Example: autism. A clinician writes: “Whenever we add T4-F8 to the mix, we often get the first word ever spoken, even at age 14, 15, and 16. Stunning to witness this.” T4-F8 is typically engaged only after two or three primary protocols.
  18. With respect to the concept of placebo washout, it should be observed that most adult clients arrive at our door after years of failed conventional treatments, all of which present opportunities for the vaunted placebo to manifest. Somehow, the placebo in these cases is temporally and spatially co-located with the neurofeedback instrument.
  19. With respect to many children’s issues, the trainees themselves may very well not know that there is a problem to be resolved. Hence there cannot be an issue of wish fulfillment. That goes against the assumptions of the placebo model.
  20. Sometimes good things happen that may even go counter to the client’s personal wishes. For example, a client with schizophrenia stops smoking in four sessions, although he never intended for that to happen.
  21. This is even more strongly the case with suicidality. In this instance, the patient’s desire is to end his life, so recovery from suicidality is not aligned with the client’s intentionality. In these cases, there may be neither the expectation of—nor the desire for—recovery. And yet it routinely happens, and often quite promptly. That cannot be assigned to placebo.
  22. In Scandinavia, a teenage girl who was made to undergo the training by her parents for her heroin addiction threatened the practitioner with a lawsuit because the training had ruined her enjoyment of the heroin. Responded the therapist: “Oh, please sue me.” He could already savor the resulting publicity.
  23. The clinical world presents many opportunities for inadvertent placebo trials in a natural context. A practitioner group that had acquired the system called somewhat chagrined some weeks later, saying that they just weren’t getting the results that they had expected. They had the electrodes connected incorrectly at the amplifier.
  24. Another such case: A person participating in a USC program for early Alzheimer’s was simultaneously also doing home-training on our neurofeedback system. After successfully training for several years, she eventually returned the system, saying that she was no longer benefiting. It turned out the problem was a broken electrode. Her training resumed successfully.
  25. Degenerative conditions such as dementia offer a particularly strong demonstration of neurofeedback efficacy that cannot be assigned to placebo. In application to degenerative conditions, recovery by neurofeedback is not permanent. On some time scale, regression sets in if training is not maintained at some level. This regression is substantially recoverable, however, if training is resumed promptly. Over time, the interval between sessions declines as the underlying organic deterioration proceeds. The cyclicality observed here is a definitive contradiction of the placebo model.
  26. Progress where none is to be expected: A family bringing their young child for training also brought along their ten-year-old with CP in a wheelchair. At one point, they asked: Might this help our older son? “Why don’t we try it,” came the answer. Within ten sessions, the child was out of the wheelchair. His parents said, “He had never expressed any interest in his legs.” Placebo, anyone?
  27. Doing the impossible: A Vietnam era veteran with a more than forty-year history of PTSD, and reporting that he has not had more than about two hours of sleep per night for five years, slept better after his first session, and after a few sessions was sleeping eight hours a night.
  28. Unexpected benefit: A 20-year-old client coming for only her third session reported that after returning home from the second session, she found that she no longer needed her glasses for reading. She had been using them since a concussion she experienced in sixth grade. This came as a complete surprise, as it was entirely unrelated to the objectives of the training.
  29. The learning curve: EEG neurofeedback behaves like a rehabilitation procedure. Gains may be slow in coming, but they are cumulative. This is not how placebos behave.
  30. Scaffolding: In complex symptom presentations, resolution is typically not achieved in parallel, as would be expected for a placebo, but rather sequentially. Recovery in one functional domain lays the basis for subsequent recovery in a related area. Training protocols that cannot be tolerated early on become indispensable later in the process.
  31. Comprehensiveness of effects: In conditions of profound dysregulation such as PTSD, recovery by neurofeedback extends over the entire range of psychophysiological symptoms. This includes many symptoms and dysfunctions that were not even on the mind of the client, and therefore ought not to be subject to recovery by the placebo mechanism. This includes in particular cognitive dysfunctions, where substantive progress can be quantitatively documented. Clients routinely report: “My ____ is better. Could that be the neurofeedback? I didn’t think it could affect that….”
  32. Improvement after training. Whereas it is the tendency of placebos to fade after a time, the tendency in neurofeedback is for progress to continue after the training is terminated. That is to be expected if the brain has in fact been induced to enhance its own functionality.
  33. Extreme change in functionality: Measurements of cognitive function to document progress in neurofeedback has yielded cases in which clients scoring initially at the one percentile level may end up scoring at the 90th percentile after a modest number of training sessions. That level of progress into the high performance domain cannot be assigned to placebo.
  34. Percent success: Throughout the modern history of the placebo model, a success rate on the order of 30% has been most commonly referred to. Success at, say, the 90% level would not ordinarily be assigned to placebo. A number of conditions exist for which success with neurofeedback is expected at the 90% level. This category includes migraine, episodic suicidality, and probably asthma as well.
  35. Commercial success: If clinicians were benefiting their clients in no more than 70% of their cases, they would not be able to sustain a successful practice. This level of success already exceeds expectations for the placebo.
  36. Completeness of resolution: Recently a paper was published on a case of pain post-cancer treatment. Symptom severity was tracked through neurofeedback training with the SCL-90. Symptom severity went from an initial value of 59 to an ending value of one. One! Such completeness of symptom resolution is inconsistent with a placebo model.
  37. Technical maturity: Neurofeedback has been practiced now for over fifty years. Techniques have been progressively refined, and results have gotten substantially better and have extended to a wider variety of conditions. Collectively, this kind of progress over the decades cannot be assigned to placebo, as there is no known mechanism for enhancing the placebo in such a manner.
  38. Rapidity of effects: For example, an ongoing migraine may be interrupted in only ten minutes of training. This is not to be expected for a placebo.
  39. Magnitude of effects: One MD experiencing her first training reported after the session that she had never experiencing such a degree of relaxation. She pronounced it fully equivalent to an Ativan drip. Significantly, she had gone into the training as a skeptic, without any positive expectations at all.
  40. Uniqueness of outcomes. Neurofeedback has achieved isolated outcomes that are outside the range of ordinary experience, and hence also outside the range of likely placebo mechanisms. For example, on a number of occasions increases in measured IQ of the order of 40 points has been achieved within a modest number of training sessions. One child saw an improvement from 70 to 112, going from developmentally delayed to above normal. Such results are unprecedented. They cannot be assigned to placebo.
  41. In every study of standard neurofeedback that has tracked IQ but one, there has been an improvement in measured IQ beyond test-retest uncertainty. There is no placebo model for this. The one exception also saw improvement, but it was only four points.
  42. Surreptitious training has effects. Sometimes brain training occurs even in the absence of any awareness that it is happening. An illustrative case: Demonstrating the system at a conference to a knowledgeable attendee who told me that he could not tolerate lateralized training in the SMR-beta range, we nevertheless hooked him up to demonstrate the system. Then we just got into conversation. A few minutes in, he suddenly became nauseous, ripped the electrodes off and dashed off to the men’s room. To my chagrin, I realized that the instrumentation had been kept running, and he had the visual display within his field of view. He had also been hearing the auditory feedback. So he had been training in background for several minutes—a case of covert training. There is no placebo model for this.
  43. Real-time effects. A woman with ptosis had her condition remediate substantially within the course of a half-hour session. The improvement was transient, but was then reinforced in subsequent sessions. A clinician who had been trying unsuccessfully to eliminate her own cluster headaches at one point moved the electrodes to an unconventional placement, and the headache disappeared over a period of ten minutes. This was the second day of an expected five-day cluster. The headache did not return on the third day…
  44. Repeatability: Working with chronic pain patients of long-standing, a fairly standard pattern emerges of a favorable response to the training, followed by a gradual re-escalation of the pain. The cycle is then repeated. Quite typically, the pain reduction is accompanied by significant change in the EEG parameters over the session. The correlated change of EEG parameters and symptom severity argues against a placebo mechanism.
  45. If a sham control is deemed to be a valid way of testing the placebo hypothesis of neurofeedback efficacy, then in fact every session of ILF training actually already fills the requirement with respect to the ILF component. The training procedure is entirely equivalent to what would be regarded as a valid sham training design. After all, the training is entirely covert. So our approach to ILF neurofeedback is indistinguishable from a sham design. And yet it has frequency-specific effects.
  46. Neurofeedback efficacy is supported by fundamental physiological measurements using fMRI, QEEG determinations, ERP changes, CNV changes, and changes in the spectral distribution in the ILF region. Changes within a single session have been documented using all of these methods.
  47. Collective experience: Nearly all clinicians who currently practice neurofeedback adopted it in mid-career. Almost none were academically prepared for that choice. All had based their professional success to date on other therapies. Yet most of those who choose to adopt neurofeedback also tend to stick with it. This means that they find neurofeedback additive to whatever they were doing before. A large minority find that it takes over their practice.
  48. Long-term training. Parents of autistic children sometimes rent the instrumentation in order to continue the training at home. This may go on for years, and calls for monthly payments. Is it reasonable to suppose that these parents are still beguiled by the placebo even after years of training? Or would it be more reasonable to suppose that they are observing actual progress in their child that is discernibly above and beyond expected developmental trajectories?
  49. Many benefits of NF can be equally well achieved with low-level auditory, visual, or electromagnetic stimulation—in a manner that emulates neurofeedback protocols. That’s an active process, which calls for no volitional or other involvement of the client. A non-placebo mechanism for the claimed effects of neurofeedback therefore must exist. A common mechanism for both stimulation and the neurofeedback training model is likely at issue. Parsimony suggests that an actual mechanism is preferable to none. The placebo hypothesis is, after all, “model-free.”
  50. All clinicians first encounter neurofeedback as skeptics. Life has already disabused them of the notion of magic remedies beyond their dreams. All bring something like the placebo model to the experience, a barrier to acceptance to be overcome. Here is one clinician’s experience, having been drawn to the field by virtue of the neurofeedback experience of her son: “[My] first experience at the introductory training was as a “volunteer” for Sue Othmer to demonstrate on; I was able to give her immediate feedback as to whether or not I was getting a headache, or feeling drowsy, all within 15 minutes [as training parameters were being optimized for her]. Any fears that I had built up memories [of my son’s training success] into something ‘too good to be true’ dissipated in those 15 minutes, and I was completely on board at that point.”
  51. Some clinicians have been doing this work for over twenty years. They have personally experienced the evolution of the method into one that has been ever so much more effective than what was available twenty years earlier. It is obviously the method that is becoming more effective over time, not the placebo.
  52. At the same time, these same clinicians have also noted increasing demands on their own competence in supervising the training. Such competence relates to the protocol specifics, the tailoring of the training parameters to each client. The entirety of the thrust toward individualization of protocols had to be evidence-based. The daily lived experience of the neurofeedback clinician therefore stands as a continual contradiction of the placebo hypothesis of neurofeedback.
  53. For the last number of years we have seen a number of former clients coming back, bringing us their children. “It’s their turn now.” Clearly, this was motivated by the memory of their own successful training experience some twenty or so years earlier.
  54. Similarly, we have seen migraine clients over the years who have come back for booster sessions. They may have suffered an injury, illness, or psychological trauma that brought about a recurrence of migraines or panic. On each occasion, they are brought to a higher level of self-regulation than before, by virtue of the intervening improvement in protocols.
  55. We have not run out of arguments yet: Clinicians in our network tend to continue the training on themselves, as do our employees. Clearly they are seeing ongoing benefit. No case can be made that such ongoing training continues to benefit from some sort of nebulous placebo. The training will have become routine.

Leave a Reply