Clever New Scheme

By Neuroskeptic | July 21, 2010 8:40 am

CNS Response are a California-based company who offer a high-tech new approach to the personalized treatment of depression: “referenced EEG” (rEEG).

This is not to be confused with qEEG, which I have written about previously. What is rEEG? It involves taking an EEG recording of resting brain activity and sending it – along with a cheque, naturally – to CNS Response, who compare it to their database of over 1,800 psychiatric patients who likewise had EEGs taken before they started on various drugs. They look to see which drugs worked best in people with an EEG profile similar to yours, and give you a fancy report with their recommendations.

That’s not completely implausible. It could work. Does it? CNS Response and some academic collaborators have just published a paper saying yes: The use of referenced-EEG (rEEG) in assisting medication selection for the treatment of depression. How solid is it? Well, it would be wrong to say that there are many problems with this study. But then if you run off a cliff and plummet into a volcano, you’ve only made one mistake.

Depressed patients were randomized to one of two groups: treatment-as-usual, which generally meant the common antidepressants bupropion, citalopram, or venlafaxine, vs. rEEG-guided personalized drug treatment. The trial was pretty large, with 114 patients randomized, and pretty long, 12 weeks. The patients had failed to respond to at least one antidepressant (mean: 1.5) during the current episode, so they were slightly “treatment-resistant”, though not extremely so.

What happened? The rEEG-guided group did better on the QIDS16SR self-report scale, and on most other measures. Not enormously: take a look at the graph, notice that the vertical axis doesn’t start at zero. But better.
Great, they did better. But why? The problem with this study is that the rEEG-guided group got a very different set of drugs to the control group. No less than 55% of them got stimulants, either methylphenidate (Ritalin) and dexamphetamine (speed). These drugs make you feel good. That’s why they’re illegal, that’s why people pay good money for them on the street.

It’s debatable whether stimulants are clinically useful as antidepressants in the long term, but they’ve got a good chance of making you feel nice for a few weeks, and make you say you feel better on a rating scale. Plus there’s nothing like a pep pill to drive active placebo effects.

The authors say that “Almost all of the studies with depression not associated with medical disorders have reported minimal or no antidepressant effect of stimulants”, and refer to some 1980s studies – yet their own trial has just shown that they do work in more than 50% of patients, and the latest Cochrane meta-analysis finds stimulants do work in the short term…

The other big names in the EEG group were MAOis (selegiline or tranylcypromine). These are often effective in treatment-resistant depression. Not necessarily more so than other drugs, but remember that these patients had already failed at least one SSRI(*). Yet the control group were, it seems, almost all given SSRIs – either citalopram, or venlafaxine, which is effectively an SSRI at low doses, e.g. the average dose used here, 141 mg. (It does other stuff, but only at higher doses of 225 mg or 300 mg.)

In summary, there were two groups in this trial and they got entirely different sets of drugs. One group also got rEEG-based treatment personalization. That group did better, but that might have nothing to do with the rEEG: they might have done equally well if they’d just been assigned to stimulants or MAOis etc. by flipping a coin. We cannot tell, from these data, whether rEEG offered any benefits at all.

What’s curious is that it would have been very simple to avoid this issue. Just give everyone rEEG, but shuffle the assignments in the control group, so that everyone was guided by someone else’s EEG. So you’d give control Patient 2 the drugs that Patient 1 should have got, and vice versa; swap 3 and 4, 5 and 6, etc.

This would be a genuinely controlled test of the personalized rEEG system, because both groups would get the same kinds of drugs. It would have been a lot easier too. For one thing it wouldn’t require the additional step of deciding what drugs to give the control group. The authors decided to follow the STAR*D treatment protocol in this study, which is not unreasonable, but that must have been a bit of a hard decision.

Second, it would allow the trial to be double-blind: in this study the investigators knew which group people were in, because it was obvious from the drug choice. Thirdly, it wouldn’t have meant they had to exclude people whose rEEG recommended they get the same treatment that they would have got in the control group… and so on.

Hmm. Mysterious. Anyway, we may be hearing more about CNS Response soon, so watch this space.

(*) – Technically, some of them had failed an SSRI and some had failed “2 or more classes of antidepressants”, but one of those classes will almost certainly have been an SSRI, because they’re the first-line treatment.

ResearchBlogging.orgDeBattista, C., Kinrys, G., Hoffman, D., Goldstein, C., Zajecka, J., Kocsis, J., Teicher, M., Potkin, S., Preda, A., & Multani, G. (2010). The use of referenced-EEG (rEEG) in assisting medication selection for the treatment of depression Journal of Psychiatric Research DOI: 10.1016/j.jpsychires.2010.05.009

  • Lemniscate

    I guess they were more interested in getting a graph published that showed 'better' results compared to the usual treatment.

  • Neuroskeptic

    I couldn't possibly comment on what their motivations may or may not have been. Cough.

  • Anonymous

    clever new scam, more like. you're too polite.

  • spike

    1 questions
    with this method, any new patient would get 'exactly' the same dosage as the ones whom their dataset is already available.right?
    if the answer is yes, then this means that there is not any new predictions(extrapolations of the previous results) which means that the effects of the drugs would be known which would preclude the huge individual differences seen in electrophysiological signals.
    so the answer should be yes! now, again if they are extrapolating or something, this means that they know the precise effects of the drugs on the features of the EEG signal, this seems very unlikely as well.

  • saltedlithium

    What would be the point of including MAOis but not MARXis?

  • Neuroskeptic

    spike: I don't think they care about the effects of drugs on EEG. All they do, as far as I can see, is to recommend drugs which worked well in people with a similar pre-treatment EEG. (and the pre-treatment EEG is taken when you're unmedicated.)

  • dearieme

    Well now, here's the sort of thing on which You Might Have Opinions?

  • Anonymous

    What's with the Journal of Psychiatric Research that published this obviously flawed study? Don't they have a peer review system? Or is this one of those corrupt journals secretly published by drug or medical device companies?

  • Neuroskeptic

    I'm not sure. It's not a leading journal, but it's not an obscure one. It has an impact factor of 3.723 so pretty middle-of-the-road.

    But the thing is, a properly designed study that found a positive effect of rEEG would be expected to go straight into the best journals in the land, i.e. the American Journal of Psychiatry probably.

  • Bernard Carroll

    Good work exposing this wishful thinking. Speaking of which, Stanford-based DeBattista and his chairman Alan Schatzberg have authored several other wishful thinking items promoting the prospects of the drug mifepristone (Corlux) for psychotic depression. Schatzberg’s company Corcept Therapeutics has the use patent for Corlux, and DeBattista was for about 2 years the medical officer for the company. The poor design of the present study of reference EEG matches the poor design of the Corcept studies. Oh, and by the way, Schatzberg is the US editor in chief of the journal in question here. Can you say editorial nepotism?

  • spike

    exactly, we are on the same line I think!
    you said:”recommend drugs which worked well in people with a similar pre-treatment EEG”
    That's exactly what I thought but my question was on the practical side of the process without getting into the detailed methods.
    So you see the pre-treatment EEG, find the most similar pre-treatment EEG in the patient dataset and see what has worked fine.
    You can clearly see that this depends heavily on the choice of features. They should be very good representatives. All sounds a bit impractical to me I must say…

  • Anonymous

    Geez, look at this monstrous conflict-of-interest disclosure…

  • Neuroskeptic

    Results section (excluding tables): 1061 words.

    COI Statement: 1084 words.

    That is never a good sign.

  • Anonymous

    Thank God Dr. Bernard Carroll is still on the watch!

  • pj

    “It's debatable whether stimulants are clinically useful as antidepressants in the long term”

    But we'll likely not know for a long time becauase of the massive resistance to their use. I've found them very useful in palliative patients (about the only group where you're allowed to give stimulants for depression). It seems to me that their abuse potential has lead to a serious resistance to even studying their efficacy (a similar argument could be made for benzos).

  • Neuroskeptic

    So you're allowed to give them to ADHD kids, but not depressed adults?

  • pj

    Pretty much.

  • Bradley Engwall, MD

    This comment has been removed by the author.

  • Bradley Engwall, MD

    This is quite fierce criticism for a pretty good paper. I would hope that critics would read papers with a modicum of care before commenting, especially academics and those with positions of professional training and responsibility.

    This journal is reputable and peer reviewed. The study was well done. The comparator group is NOT “treatment-as-usual”, but rather treatment acording to STAR*D algorithm which is state of the art for clinical psychiatry. The fact that the STAR*D algorithm doesn't include stimulants is not the fault of the study design, rather it reflects prevailing expert opinion and the state of the psychiatric literature (few trials of stimulants for depression, esp. long term).

    Additionally, I don't think we can conclude that stimulants *aren't* effective from our current literature. The data just isn't there.

    My perspective as a pracititioner who has some experience with rEEG, though not controlled in any randomized or quantitative fashion, is that it's an imperfect, but remarkable tool for patients resisitant to the usual treatments for psychiatric disorders. With a larger database of medications and supplements rEEG holds the promise of the psychiatric Shangri-La: an objective tool to guide prescribing practice. While I mourn the fact that it is not an open source tool, the cost is extremely reasonable compared to doctors fees and the cost of medication. Time will tell if it's usefulness is borne out. In my practice it is.

    (Just so all are aware of my interests: I own no stock and have no financial relationships to anyone in the company and was not asked to provide this opinion. (They list me as a provider on my website, but I generally don't accept medication only referrals: usually preferring to do psychotherapy with or without medication as appropriate.) I'm just passionate about good clinical care.)

  • Neuroskeptic

    Bradley Engwall: Thanks for the comment.

    Whether STAR*D is state of the art is debatable. Personally, if you have someone who is known to be resistant to 1 or more SSRIs, I don't think that giving them venlafaxine at a dose (<150 mg) at which it acts solely as an SSRI, is very clever.

    But I'm not a clinician.

    However, that's by the by, the point is that this wasn't a randomized controlled trial of rEEG because, like I said, the drugs used in the two groups differed.

    The rEEG group did better, but for all we know, they'd have done equally as well if they'd just been given stimulants & MAOis, forget the EEG.

    You're right in a sense that we can't blame rEEG for the fact that STAR*D doesn't include stimulants. But that just supports my point. Maybe all we need to do is use more stimulants. Not use more stimulants guided by rEEG.

    As I said, the only way to test rEEG would be to randomize people to 1) real rEEG or 2) the “wrong” rEEG (guided by someone else's EEG).



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar