Research - Placebo controlled double-blind studies with EEG Biofeedback - Methods |
|
|
The following discusses methods to produce a placebo EEG biofeedback experience for clinical research. It describes some approaches that have been evaluated, and used, in the design of such experiments. The intent is that, the same way that a pharmaceutical research study can use either the real medicine or a sugar pill, and the subjects do not know the difference, we want to provide either real EEG biofeedback or "sham" EEG biofeedback to subjects. The purpose is to prove that the real EEG biofeedback really works, and that its benefits are not just due to the experience of sitting in front of a computer, thinking you are seeing your brainwaves, expecting it to work, etc.
The goal is to provide a basis to perform a double-blinded, placebo-controlled study of EEG biofeedback. There will be two matched groups. The experimental group will receive EEG biofeedback training as a theraputic intervention. The control group will receive an experience that is a placebo for this treatment. This necessitates providing, as a control condition, an experience, for both the subject and the investigator, that is identical to real EEG biofeedback, yet does not have the proper operant conditioning cues. In other words, the experience is indistinguishable from EEG biofeedback, yet learning is not able to occur, because the experience lacks the proper contingency relationship to the subject's EEG. There are a variety of ways to provide such an EEG placebo.
One method is to deliver a random signal ("simulation") in lieu of real eeg. The simulated EEG is a random signal that has frequency content and time behavior similar to EEG. It looks somewhat like EEG, and it can provide essentially similar time behavior of feedback such as tones, videos, etc. The aggregate rate of feedback can be set at whatever level is desired, to simulate a session with as much or as little reward as desired. The trainee would be experiencing, instead of their own EEG, a signal that the computer generates, which is designed to seem like EEG to the untrained observer. Such a simulation does not, however, have statistics and subtle time behavior that is identical with EEG. Rather, it tends to be more monotonous, providing a steady amount of feedback at all times. This potential monotony, and the fact that the signal has subtely different statistics from real EEG, may be a drawback for this design. Increasingly complex statistics, including time-varying statistics can be employed. However, the challenge remains to demonstrate that the placebo is "in all ways" identical to real EEG, and this burden may be an unending one.
A second, and somewhat preferable method, is to use played-back real EEG, which is played back like a tape recorder, internal to the software. The EEG would have been gathered previously from a different session, stored digitally on the computer using the BrainMaster software, and then played back through the software, in real time. The playback has the same time behavior of the live EEG, including how many screens are updated per second, how the signal processing handles the data, etc. Again, the appearance of the onscreen data and training screens is indistinguishable from real EEG. The main advantage of this approach is that the subject is experiencing true EEG in time and in feedback, so that any subtle dynamics or variations that occur in real EEG will be present in the placebo. In other words, the "experience" will be a true EEG experiences.
Therefore, the use of playback of real EEG is preferable, in that it provides a more realistic EEG experience for the sham operation, and minimizes differences between the control and the experimental condition. The played-back EEG can be that of the same individual from a prior session, or it can be EEG from another person. One drawback of playing back one's own EEG is that one might conceivable recognize patterns in it, and thus recognize that it is not real time. Therefore, it may be desirable to cross-match subjects, so that a given control subject is presented with EEG that was gathered from either another control subject, or even from an experimental group subject. By playing back EEG from a different individual, it is ensured that there is no way the subject could recognize the experience by recalling any details of the feedback.
A possible drawback with both of these methods is that the trainee would not experience the effect of artifact, such as motion, eyeblinks, etc. Care will need to be taken so that this difference is not detected. The subjects will need to be instructed to sit still, and if they move or do anything that would produce artifact, this event should not be noted in either the control or the experimental group. If this becomes an issue, then a custom software method could be devised, in which either simulated or played-back EEG is combined with the users's own EEG, so that artifacts do make it into the signal, adding this level of realism. There would be cost and time associated with implementing this alternative, however, and it is likely not necessary, with a thorough design and script.
In final design of an experiment, care must be taken to consider the computer displays and messages available to the operator, the subject, and any subject "coaches". At various points, the software "knows" whether it is playing real EEG, played-back EEG, or simulations, and the double-blind design must ensure that such revealing information is not available to any of the blinded participants. This may call for special considerations in layout of the lab, use of the software, and instructions to users. In one study design, the subject and the investigator are in one room watching the "feedback" screens, while the actual control occurs in another room, by a person who is not blinded to the conditions, but who is blinded to the identities of the participants, and is also uninvolved in the study, and is merely an "operator".
One consideration with any of these approaches is the phenomenon of "learned helplessness" in control subjects who experience many sessions of non-efficacious biofeedback. This can lead to a defeat of the learning mechanism in subsequent sessions when true feedback is applied. This is a concern for control subjects whose first experience is a certain amount of sham experience, that they believe is a form of biofeedback therapy. The subject's expectations play into this. If they are expecting to improve and do not, then the learned helplessness can occur. If the control subjects do not expect a benefit, but are merely experiencing the feedback for some other purported reason, then learned helplessness can be reduced or eliminated.
Any of these methods can be applied equally well to high-frequency, eyes-open protocols such as alpha, beta, or SMR training, and to low-frequency eyes-closed methods such as Penniston alpha/theta training.
|
|
Powered by
KBPublisher (Knowledge base software)