TCPR: You and your UMass colleagues recently reviewed the current research on the efficacy of neurofeedback in psychiatry (Begemann M et al, Trans Brain Rhyth 2016;1(1):19–29). But before going into the research, can you please describe what the procedure is and how it is done? Many of our readers wonder whether neurofeedback is a legitimate or effective treatment, and whether it might help our patients with ADHD.
Dr. van Lutterveld: Neurofeedback is a type of biofeedback that uses an EEG to provide feedback to patients. While hooked up to an EEG, patients are shown different images through various forms of media, including video and other tools. The general idea is that the EEG can detect brainwaves that are associated with improvement in various symptoms, such as inattention in ADHD. The patient is then taught how to produce more of those “healthier” brain waves, theoretically leading to improvements in symptoms.
TCPR: Can you give us an example of what a participant would actually see when using an EEG?
Dr. van Lutterveld: It depends on the setup being used. Some systems will provide the feedback in the form of pretty landscapes, and as people successfully alter their brain waves in the appropriate direction, the landscape will change, going from winter to spring to summer, and then deer and rabbits start to appear. In those cases, it’s really nice. Often, though, and especially in recent studies, there hasn’t been much energy put in to make it look nice for the participants. So, what they see can also be as simple as just a graph.
TCPR: So, the participants are told by the researcher or the clinician to look at a monitor and to try to change the image in some way, and certain changes are correlated with certain EEG changes?
Dr. van Lutterveld: Yes. It’s brain training, and the participants get immediate feedback on how they are doing through the visual presentation.
TCPR: How many sessions are usually required?
Dr. van Lutterveld: That can vary, but it’s usually between 20 and 50 sessions, and the entire treatment can take from 1 to 2 months.
TCPR: And these sessions would be daily—like a 5-day-a-week kind of thing?
Dr. van Lutterveld: It basically depends on the study itself, but usually it’s done once a day and then for several weeks. But you can also have people come in twice a week and then prolong it. Again, it’s an open question; we don’t know right now, and we need more research on what’s going to work best.
TCPR: Are the EEGs used in neurofeedback the same as the standard EEGs used in neurology?
Dr. van Lutterveld: Yes, but in neurofeedback typically only one electrode is used, whereas in neurology—for example, to diagnose seizures—many electrodes are used. When you are measuring brain activity to make a diagnosis, generally the more electrodes the better. But when providing neurofeedback, you look at how strong a specific brain wave is in a single electrode. You then provide patients with a visual measure of how strong the wave is and teach them how they can upregulate it or downregulate it. (See page 1 for a primer on using EEG in psychiatry.)
TCPR: Is one electrode good enough?
Dr. van Lutterveld: Probably not, but because it takes much more time to place them, neurofeedback doesn’t typically use multiple electrodes. EEG has very good temporal resolution, but it does not have good spatial resolution. This means that EEG can measure brain activity at the millisecond level, even the sub-millisecond level, and it’s very fast. But to localize the source of the waves, you need many electrodes, usually in the range of 60 or more. Using only one electrode makes it difficult to determine where the signal is coming from, and this is one of the methodological problems with EEG neurofeedback: You don’t know which brain region you are actually training.
TCPR: When are the brain waves generated?
Dr. van Lutterveld: Basically, they originate when a certain number of neurons fire in a specific frequency, and the signal strength from this group of neurons must be strong enough to travel through the skull and the scalp. And there are different frequency bands, which are associated with different cognitive functions. It’s not only the wave frequency that is important, but the location in the brain from where the wave comes.
TCPR: Can you give us an example of why knowing the brain location is important?
Dr. van Lutterveld: Sure, here’s one. The EEG theta waves are very important in the hippocampus, and they are related to memory processes. But theta waves from the frontal brain regions are associated with focus and attention, deep relaxation, and meditation.
TCPR: Are there certain brain waves that are more associated theoretically with ADHD?
Dr. van Lutterveld: When it comes to the treatment of children and adolescents with ADHD, in the literature researchers often look at beta and theta waves. The electrode measures a complete spectrum of waves called the broad band, and that signal is sent to the neurofeedback machine, which basically extracts the beta and theta waves. The machine calculates wave strength, and the goal is to then enhance beta waves and suppress theta waves (Duric NS et al, BMC Psychiatry 2012;12:107).
TCPR: So, now back to the big question: What’s the evidence that this works?
Dr. van Lutterveld: In our study (which was led by Dr. Marieke Begemann), we did a database search to find 30 studies that met our criteria for good methodology. There were 19 studies for ADHD, 5 for autism spectrum disorder, 3 for OCD, 2 for generalized anxiety disorder, and 1 for depression. The oldest study was published in 1995, but most were far more recent, with several published in 2013 or 2014 (Christiansen H et al, Front Hum Neurosci 2014;8:943).
TCPR: And what were the methodological criteria? Did they have to be double-blind, randomized controlled trials?
Dr. van Lutterveld: No, we were much more lenient, because otherwise we would have only found 3 trials to include. So, we required only that the studies have some type of control group to compare with the treatment group, and that they report some type of outcome measure.
TCPR: And what did you find?
Dr. van Lutterveld: Mainly, we found that because the studies had so many deficiencies, it was impossible to conclude very much.
TCPR: Can you expand on that?
Dr. van Lutterveld: Most of them were open-label, which basically means that the treatment being administered was known to the researchers, which can lead to bias. Also, many studies compared neurofeedback with “treatment as usual.” Because you want to have a control group that is as similar as possible to the treatment group, that’s not good enough. The control group should get sham neurofeedback, or at the least an EEG. We found that most studies that used a waiting list control group, or other type of non-comparable control group, reported some benefit of neurofeedback. But the only study that used a good sham control group found no benefit. In addition, several studies were not randomized, which limits the validity of the results. Other methodological concerns included a small sample size, a lack of standardization on what the type of neurofeedback provided, and a lack of reporting on how many people responded and how many people didn’t.
TCPR: If they didn’t report on how many people responded, what data did they report?
Dr. van Lutterveld: They usually reported scores on a symptom rating scale. So, theoretically, you might see that the experimental group shows a 2% decrease in inattention vs the control group, and even if that is statistically significant, clinically it is generally not significant. The more clinically useful outcome would be to set a threshold that is defined as a response, such as reducing symptoms to a certain level on a rating scale, and reporting the percentage of people responding in the neurofeedback vs the control group. But most of the studies did not report that.
TCPR: You mention small sample sizes.
Dr. van Lutterveld: The sample sizes were mostly very small. For example, for OCD, we included a study that had 4 participants per group. If you want to detect a medium effect, you need a minimal sample size of 64 per group. We found no study that met that requirement.
TCPR: And how does a small sample size limit what we can conclude from the studies?
Dr. van Lutterveld: You have low statistical power. Underpowered studies are problematic in two ways. If they fail to show a benefit of a treatment, that may be a false negative; a similar study with more people might uncover a difference. But on the other hand, if you do find a difference with small samples, you don’t know if that difference actually reflects reality or is simply a chance finding that you are seeing in this particular small group of people. The other problem with small samples is a bit subtler and is a kind of a publication bias. If I do an underpowered study with a small sample, and I find no effect and I submit the study for publication, chances are the study will be rejected by the editors. They will say, “How can you know that there is no effect? If you increase your sample size, you will increase your statistical power, and maybe then you will find an effect.”
TCPR: Why is that?
Dr. van Lutterveld: Because the peer reviewers will criticize the study by saying that they didn’t find a difference because the sample size was too small. So, these papers are often rejected. But when the same small sample yields a positive result, that paper is more likely to be published. In the end, the literature is artificially enriched with these “positive” studies that have small samples and therefore may not really generalize to the population as a whole.
TCPR: Given that the research evidence is so unimpressive, why is there still so much interest in neurofeedback?
Dr. van Lutterveld: One of the reasons people are interested in neurofeedback is that the side effect profile is minimal. This is true, but there is a cost to the treatment. Patients are going to treatment every day for 1 or 2 months, which translates into thousands of dollars per month, depending on the clinic. In addition, patients may be wasting time on an unproven treatment, delaying more established treatment options.
TCPR: I know your lab at UMass is working on neurofeedback research. Can you describe what that research is? Does it overcome some of the methodological challenges we’ve been discussing?
Dr. van Lutterveld: We are looking at whether neurofeedback can help people develop mindfulness, and whether this can help them deal with stress and anxiety. We started by using functional MRI to pinpoint a specific brain region called the posterior cingulate cortex (PCC), which is related to mindfulness. To do this, we put people in an fMRI scanner and used the fMRI signal as neurofeedback.
TCPR: What did you find?
Dr. van Lutterveld: We found a very strong correlation between the subjective experience of mindfulness and a decreased PCC signal. The next step was replicating that with EEG neurofeedback. In this case, we used many leads so that we could localize the brain waves to the PCC area, and we again found a good correlation between mindfulness and gamma waves.
TCPR: And what’s the next step in the research?
Dr. van Lutterveld: Now we are testing it in a randomized control trial, where we are randomly assigning people to EEG neurofeedback vs sham EEG. Both groups are given an 8-week mindfulness course, but only the treatment group is also given real neurofeedback. We hope to demonstrate that those receiving neurofeedback develop more mindfulness and decrease their PCC activity.
TCPR: Before we conclude, do you have any other thoughts about neurofeedback?
Dr. van Lutterveld: My take-home message is that, if neurofeedback works, it would be great, especially nowadays when people can use neurofeedback with smartphone headsets that cost around $100. But it’s important to first establish sound research showing that it actually works. To do that, we need studies that are well-done, well-performed, and methodologically sound.
TCPR: Thank you for your time, Dr. van Lutterveld.