Serotonin, Psychedelics and Depression

Note: This post is part of a Nature Blog Focus on hallucinogenic drugs in medicine and mental health, inspired by a recent Nature Reviews Neuroscience paper, The neurobiology of psychedelic drugs: implications for the treatment of mood disorders, by Franz Vollenweider & Michael Kometer. That article will be available, free (once you register), until September 23. For more information on this Blog Focus, see the "Table of Contents" here.

Neurophilosophy is covering the history of psychedelic psychiatry, while Mind Hacks provides a personal look at one particular drug, DMT. The Neurocritic discusses ketamine, an anesthetic with hallucinogenic properties, which is attracting a lot of interest at the moment as a treatment for depression.

Ketamine, however, is not a "classical" psychedelic like the drugs that gave the 60s its unique flavor and left us with psychedelic rock, acid house and colorful artwork. Classical psychedelics are the focus of this post.

The best known are LSD ("acid"), mescaline, found in the peyote and a few other species of cactus, and psilocybin, from "magic" mushrooms of the Psilocybe genus. Yet there are literally hundreds of related compounds. Most of them are described in loving detail in the two heroic epics of psychopharmacology, PIKHaL and TIKHaL, written by chemists and trip veterans Alexander and Ann Shulgin.

The chemistry of psychedelics is closely linked with that of depression and antidepressants. All classical psychedelics are 5HT2A receptor agonists. Most of them have other effects on the brain as well, which contribute to the unique effects of each drug, but 5HT2A agonism is what they all have in common.

5HT2A receptors are excitatory receptors expressed throughout the brain, and are especially dense in the key pyramidal cells of the cerebral cortex. They're normally activated by serotonin (5HT), which is the neurotransmitter that's most often thought of as being implicated in depression. The relationship between 5HT and mood is very complicated, and depression isn't simply a disorder of "low serotonin", but there's strong evidence that it is involved.

There's one messy detail, which is that not quite all 5HT2A agonists are hallucinogenic. Lisuride, a drug used in Parkinson's disease, is closely related to LSD, and is a strong 5HT2A agonist, but it has no psychedelic effects. It's recently been shown that LSD and lisuride have different molecular effects on cortical cells, even though they act on the same receptor - in other words, there's more to 5HT2A than simply turning it "on" and "off".

*

How could psychedelics help to treat mental illness? On the face of it, the acute effects of these drugs - hallucinations, altered thought processes and emotions - sound rather like the symptoms of mental illness themselves, and indeed psychedelics have been referred to as "psychotomimetic" - mimicking psychosis.

There are two schools of thought here: psychological and neurobiological.

The psychological approach ruled the first wave of psychedelic psychiatry, in the 50s and 60s. Psychiatry, especially in America, was dominated by Freudian theories of the unconscious. On this view, mental illness was a product of conflicts between unconscious desires and the conscious mind. The symptoms experienced by a particular patient were distressing, of course, but they also provided clues to the nature of their unconscious troubles.

It was tempting to see the action of psychedelics as a weakening of the filters which kept the unconscious, unconscious - allowing repressed material to come into awareness. The only other time this happened, according to Freud, was during dreams. That's why Freud famously called the interpretation of dreams the "royal road to the unconscious".

Psychedelics offered analysts the tantalizing prospect of confronting the unconscious face-to-face, while awake, instead of having to rely on the patient's memory of their previous dreams. To enthusiastic Freudians, this promised to revolutionize therapy, in the same way that the x-ray had done so much for surgery. The "dreamlike" nature of many aspects of the psychedelic experience seemed to confirm this.

Not all psychedelic therapists were orthodox Freudians, however. There were plenty of other theories in circulation, many of them inspired by the theorists' own drug experiences. Stanislav Grof, Timothy Leary and others saw the psychedelic state of consciousness as the key to attaining spiritual, philosophical and even mystical insights, whether one was "ill" or "healthy" - and indeed, they often said that mental "illness" was itself a potential source of spiritual growth.

Like many things, psychiatry has changed since the 60s. Psychotherapy is currently dominated by cognitive-behavioural (CBT) theory, and Freudian ideas have gone distinctly out of fashion. It remains to be seen what CBT would make of LSD, but the basic idea - that carefully controlled use of drugs could help patients to "break through" psychological barriers to treatment - seems likely to remain at the heart of their continued use.

*

The other view is that these drugs could have direct biological effects which lead to improvements in mood. Repeated use of LSD, for example, has been shown to rapidly induce down-regulation of 5HT2A receptors. Presumably, this is the brain's way of "compensating" for prolonged 5HT2A activation. This is probably why tolerance to the effects of psychedelics rapidly develops, something that's long been known (and regretted) by heavy users.

Vollenweider and Kometeris note that this is interesting, because 5HT2A blockers are used as antidepressants - the drugs nefazadone and mirtazapine are the best known today, but most of the older tricyclic antidepressants are also 5HT2A antagonists. Atypical antipsychotics, which are also used in depression, are potent 5HT2A antagonists as well.

So indirectly suppressing 5HT2A might be one biological mechanism by which psychedelics improve mood. However, questions remain about how far this could explain any therapeutic effects of these drugs. Psychedelic-induced 5HT2A down-regulation is presumably temporary - and if all we need to do is to knock out 5HT2A, it would surely be easiest to just use an antagonist...

Vollenweider FX, & Kometer M (2010). The neurobiology of psychedelic drugs: implications for the treatment of mood disorders. Nature Reviews Neuroscience, 11 (9), 642-51 PMID: 20717121

Cats, Bins and Stalin

Britain is currently being outraged by the woman who threw a cat in a bin for some reason, and got caught on video:

As a cat person I'm as outraged as anyone, but as a vegetarian I feel that carnivores who object to this are not being very consistent. To paraphrase something that Stalin didn't actually say:

One cat in a bin is a tragedy. 2 million chickens killed every day is delicious

The life of a broiler chicken is not a happy one.

You Read It Here First

Remember the paper from 2009 about combining two different drugs in the treatment of depression?

It was about a clinical trial in which patients were randomly assigned to get just one antidepressant, fluoxetine, or two - mirtazapine & fluoxetine, mirtazapine & venlafaxine, or mirtazapine & buproprion. The people who got two antidepressants did better.

But as I said at the time, in a comment beneath my post about it...

All the first 6 weeks shows is that mirtazapine is better than placebo. Everyone in the study got a non-mirtazapine antidepressant, so any improvement in the non-mirtazapine group (i.e. the fluoxetine alone group) could have been placebo, regression to the mean etc. The only placebo-controlled aspect was that some people got placebo mirtazapine and some people got real mirtazapine.

Now Dr's El-Mallakh, Kaur and Lippmann have written in a Letter to The Editor of the American Journal of Psychiatry (where the original paper appeared) that

There was no mirtazapine plus placebo study group. This comparison arm is necessary in order to be confident that the observed effect by the three combined treatments could not have been accomplished by mirtazapine as a single drug. The observation that mirtazapine alone was equivalent to fluoxetine or paroxetine alone in a previous study does not negate the need for a control in the Blier et al. study. Without such a control, one cannot assume that two antidepressant medications are more effective than mirtazapine alone.

What I said - on 18th December 2009. The new Letter was "accepted for publication" in May 2010, and it's only just appeared.

Am I just blowing my own trumpet? No. Well, a bit. But there's a serious point as well: internet comments are a much better medium for discussing and criticizing research than Letters To The Editor ever can be.

Why? The Letter may have been a bit slower, but it's still out there, surely? Plus, it'll have been read by far more people. My post has got about 400 pageviews so far. I don't know how many people read the Letters page in the AJP, but I'd imagine it must be a good few thousand. So what's the problem?

The problem is that it's too late. Papers get cited by other papers fast (this one's got 13 citations so far), and they change minds even faster. This article's been out nearly a year, and I'm sure that in that time it will have convinced some psychiatrists to start their depressed patients on two drugs, rather than just one.

Now I'm not saying they shouldn't do that. I don't know. Anyway, I'm not a doctor. But I stand by my comment that this paper shouldn't be what changes your opinion on that question; the design of the trial means it can't tell you that. And I think that's something that readers of the paper should have been told at the time, not 9 months later.

What's the solution? I've written about this previously as well. Scientific journals should have open, blog-style comment threads attached to everything they publish, so that readers can say what they have to say, immediately. A number of major journals, e.g. the PLoS journals, some of the Nature ones, and the BMJ, already do this.

From what I've seen, the standard of comments is extremely high. Sure, some are rubbish. But the rubbish ones are almost always obviously bad, so I don't think they'll be doing much damage. The good ones, on the other hand, are often extremely insightful - whether they are criticizing, or praising, the paper.

El-Mallakh RS, Kaur G, & Lippman S (2010). Placebo group needed for interpretation of combination trial. The American journal of psychiatry, 167 (8) PMID: 20693473

Help I'm Being Regressed To The Mean

"Regression to the mean" was the bane of my undergraduate statistics class. We knew that it was out there, and that the final exam would have a question about it, but no-one understood it or had ever seen it. A bit like unicorns or fairies.

The lecture notes were unhelpful. They told us what it did - make things wrongly appear to change over time when actually stuff stayed the same - but not what it was. Some people claimed to get it, but they couldn't explain it to others.

I now see that our mistake was in thinking that there's some thing called "regression to the mean". There isn't. It's just a rather unhelpful term for what happens in a certain kind of situation, and once you understand those situations, there's nothing more to learn.

Suppose there's a number, which varies over time, and at least some of this variation is random. It could be anything from the number of sunspots to rates of cancer. You get interested in this number whenever it gets very high (or very low). Whenever it does, you start tracking the number for a while. Maybe you even try to change it. You notice that the number always seems to be falling (or rising). Why?

Because you only get interested in the number when it's, by chance, unusually high. The chances are, the next time you look at it, it will be lower: not for any interesting reason, or because "what goes up must come down", but just because if you take an unusually high number and then generate a new number at random, it'll probably be lower. That's why the first number was "unusually high".

Suppose that you take some people and give them an IQ test twice, a week apart. Call the first test "X" and the second test "Y". Suppose it's a crap test that gives entirely random results. Here's what might happen if you gave the test to 100 people, with each dot a person:
There's no correlation, because X and Y are both random junk. Nothing to see, move along. But wait a second...
Here's X, first test score, plotted vs Y-X i.e. the change in score between the first test and the second. There's a strong negative correlation: people who did well on the first test tended to get worse, and people who did badly, tended to improve. Wow? No. This is a purely statistical effect. It's meaningless: the "correlation" exists only because we're correlating X with itself (in the form of Y-X).

It's a fundamental mistake, and it's obvious when you look at it like this, yet it's a surprisingly easy one to make without noticing. Imagine that you'd invented a pill that you think can make people smarter. You decide to test it on "stupid people", because they're the ones who need it most. So you give lots of people an IQ test (X), select the worst 10%, and give them the drug. Then you re-test them afterwards (Y). Whoa! They've improved! The drug works!

There's only one stupid person involved in this experiment.

This remains true, even if the IQ tests aren't entirely random. A test that measures real intelligence will also have an element of luck. By selecting the bottom 10% of scores, you're selecting people who are both unintelligent and unlucky when they took the test. They'd have scored 11% if they were lucky. So the same problem applies, albeit to a lesser degree.

That's really all there is to "regression to the mean". The regression of high or low scores towards the mean score is inevitable, given our definition of "high" and "low" scores, to the extent that scores are random. This is why I said it's unhelpful to think of it as a thing. The trick is being able to spot it when it happens, and to avoid being mislead by it. If you're not careful, it can happen anywhere.

Interestingly, the reason why it's thought of in this unhelpful way is probably because the "discoverer" of regression-to-the-mean, Francis Galton, misunderstood it. He observed this "effect" in some data he'd collected about human height, and he wrongly interpreted it as a real biological fact about genetics. Eventually, people noticed the statistical mistake, but the idea of "regression to the mean" stuck, to the dismay of undergraduates everywhere.

Link: This was inspired by a post on Dorothy Bishop's blog, Three ways to improve cognitive test scores without intervention.

Fish Out Of Water, On Ketamine

Ketamine is a drug of many talents. Used medically as an anesthetic in animals and, sometimes, in humans, it's also become widely used recreationally despite, or perhaps because of, its reputation as a "horse tranquilizer".

Ketamine's also a hot topic in research at the moment for two reasons: it's considered an interesting way of provoking the symptoms of schizophrenia, and it's also shown promise as a fast-acting antidepressant.

Anyway, most ketamine research to date has been done in either humans or in rodents, but New York pharmacologists Zakhary et al decided to see what it does to fish. So they put some ketamine in the fishes water and saw what happened: A Behavioral and Molecular Analysis of Ketamine in Zebrafish.

A high dose, 0.8%, just made the fish unconscious. Well, it is an anesthetic. But a low dose (0.2%) had rather more complex effects. It sent them literally loopy - they started swimming around and around in circles, usually in a clockwise direction. Control zebrafish swam about and explored their tanks without any circling behaviours.

They also examined the effect of ketamine on the "hypoxic stress" response, i.e. what happens when you take the fish out of water (only for 20 seconds, so it doesn't cause any real harm.) Normal fish struggle and gasp for water in this situation, unsurprisingly. Ketamine strongly inhibited this.

So what? Well, it's hard to say what this might mean. It would be great if the zebrafish turned out to be a useful experimental model for investigating the effects of ketamine and similar drugs, because they're much easier to work with than rodents (for one thing, it's a lot easier to just put a drug in a fish tank than to inject it into a mouse.)

However, it remains to be seen whether swimming in circles is a useful analog of the human effects of ketamine. Ketamine can make people act in some pretty stupid ways, but walking around in little circles is extreme even by K-head standards...

Link: I've blogged about ketamine before: I'm On K, You're On K.

Zakhary SM, Ayubcha D, Ansari F, Kamran K, Karim M, Leheste JR, Horowitz JM, & Torres G (2010). A behavioral and molecular analysis of ketamine in zebrafish. Synapse (New York, N.Y.) PMID: 20623473

PR Reviewed Research

Suppose you've done some research, but unfortunately, it's crap.

Maybe your methods are flawed. Or your data don't really support the conclusions you want to draw from them.

You seem to be out of options. You could release the research, but then people would criticize it, or you could keep quiet about it, but then you've wasted all the time and money you spent on it. Neither is very attractive.

But there's a third option. Publicize the conclusions of your work, along with the best cherry-picked results, before you actually release the full report. Write a press release which "for reasons of space" only discusses the sexy stuff. You could even make it out to be a "leak", if you were feeling really devious.

Everyone will start talking about what you've said, despite the fact that without the full data, it's just your claims that you might as well have pulled out of thin air. Yet no-one can criticize it because no-one knows what your methods were. Leave it a few weeks, and then when you eventually do release the details, no-one will care any more - but the message has got out there.

*

On an unrelated note, a British management consultancy firm have done some research concluding that British local government employees work less efficiently than their counterparts in business, due to poor management. Sounds like a (very very big and lucrative) job for a management consultant!

Unfortunately, the details of the research aren't available yet. Their website tell us that

The research into public sector productivity will be available as a down load from this site when the report is released at the end of August.

But the conclusions are available right now, and are all over the media, including even the BBC whose remit must have expanded to cover advertising while I wasn't looking.

Junior staff in local authorities were, on average, productive only 32% of the time during working hours, said [the] management consultancy... It said this compared with an average of 44% in the private sector.

Is this true? We have absolutely no way of knowing because all we're told about the methodology was that it involved

1,855 surveys of managers and supervisors (173 from local government officers), 376 day-long observations, comprised of a minute by minute categorisation of how the manager in question spent his time, of which 36 were from local government.

Sounds like it could be pretty solid. Or it could be complete bollocks. The devil is in the details as it always is with research: what were the survey questions? Were the samples representative? What was the compliance rate? Were the people who did the minute-to-minute categorization of manager's time blinded to whether the manager was public or private sector?

No doubt we'll be informed as to all this in about two weeks, by which time no-one will care - but the message has got out there.

Fantastic.

Schizophrenia, Genes and Environment

Schizophrenia is generally thought of as the "most genetic" of all psychiatric disorders and in the past 10 years there have been heroic efforts to find the genes responsible for it, with not much success so far.A new study reminds us that there's more to it than genes alone: Social Risk or Genetic Liability for Psychosis? The authors decided to look at adopted children, because this is one of the best ways of disentangling genes and environment.

If you find that the children of people with schizophrenia are at an increased risk of schizophrenia (they are), that doesn't tell you whether the risk is due to genetics, or environment, because we share both with our parents. Only in adoption is the link between genes and environment broken.

Wicks et al looked at all of the kids born in Sweden and then adopted by another Swedish family, over several decades (births 1955-1984). To make sure genes and environment were independent, they excluded those who were adopted by their own relatives (i.e. grandparents), and those lived with their biological parents between the ages of 1 and 15. This is the kind of study you can only do in Scandinavia, because only those countries have accessible national records of adoptions and mental illness...

What happened? Here's a little graph I whipped up:

Brighter colors are adoptees at "genetic risk", defined as those with at least one biological parent who was hospitalized for a psychotic illness (including schizophrenia but also bipolar disorder.) The outcome measure was being hospitalized for a non-affective psychosis, meaning schizophrenia or similar conditions but not bipolar.

As you can see, rates are much higher in those with a genetic risk, but were also higher in those adopted into a less favorable environment. Parental unemployment was worst, followed by single parenthood, which was also quite bad. Living in an apartment as opposed to a house, however, had only a tiny effect.

Genetic and environmental risk also interacted. If a biological parent was mentally ill and your adopted parents were unemployed, that was really bad news.

But hang on. Adoption studies have been criticized because children don't get adopted at random (there's a story behind every adoption, and it's rarely a happy one), and also adopting families are not picked at random - you're only allowed to adopt if you can convince the authorities that you're going to be good parents.

So they also looked at the non-adopted population, i.e. everyone else in Sweden, over the same time period. The results were surprisingly similar. The hazard ratio (increased risk) in those with parental mental illness, but no adverse circumstances, was 4.5, the same as in the adoption study, 4.7.

For environment, the ratio was 1.5 for unemployment, and slightly lower for the other two. This is a bit less than in the adoption study (2.0 for unemployment). And the two risks interacted, but much less than they did in the adoption sample.

However, one big difference was that the total lifetime rate of illness was 1.8% in the adoptees and just 0.8% in the nonadoptees, despite much higher rates of unemployment etc. in the latter. Unfortunately, the authors don't discuss this odd result. It could be that adopted children have a higher risk of psychosis for whatever reason. But it could also be an artefact: rates of adoption massively declined between 1955 and 1984, so most of the adoptees were born earlier, i.e. they're older on average. That gives them more time in which to become ill.

A few more random thoughts:

• This was Sweden. Sweden is very rich and compared to most other rich countries also very egalitarian with extremely high taxes and welfare spending. In other words, no-one in Sweden is really poor. So the effects of environment might be bigger in other countries.
  
• On the other hand this study may overestimate the risk due to environment, because it looked at hospitalizations, not illness per se. Supposing that poorer people are more likely to get hospitalized, this could mean that the true effect of environment on illness is lower than it appears.
  
• The outcome measure was hospitalization for "non-affective psychosis". Only 40% of this was diagnosed as "schizophrenia". The rest will have been some kind of similar illness which didn't meet the full criteria for schizophrenia (which are quite narrow, in particular, they require >6 months of symptoms).
• Parental bipolar disorder was counted as a family history. This does make sense because we know that bipolar disorder and schizophrenia often occur in the same families (and indeed they can be hard to tell apart, many people are diagnosed with both at different times.)

Overall, though, this is a solid study and confirms that genes and environment are both relevant to psychosis. Unfortunately, almost all of the research money at the moment goes on genes, with studying environmental factors being unfashionable.

Wicks S, Hjern A, & Dalman C (2010). Social Risk or Genetic Liability for Psychosis? A Study of Children Born in Sweden and Reared by Adoptive Parents. The American journal of psychiatry PMID: 20686186

fMRI Analysis in 1000 Words

Following on from fMRI in 1000 words, which seemed to go down well, here's the next step: how to analyze the data.

There are many software packages available for fMRI analysis, such as FSL, SPM, AFNI, and BrainVoyager. The following principles, however, apply to most. The first step is pre-processing, which involves:

• Motion Correction aka Realignment – during the course of the experiment subjects often move their heads slightly; during realignment, all of the volumes are automatically adjusted to eliminate motion.
• Smoothing – all MRI signals contain some degree of random noise. During smoothing, the image of the whole brain is blurred. This tends to smooth out random fluctuations. The degree of smoothing is given by the “Full Width to Half Maximum” (FWHM) of the smoother. Between 5 and 8 mm is most common.
• Spatial Normalization aka Warping – Everyone’s brain has a unique shape and size. In order to compare activations between two or more people, you need to eliminate these differences. Each subject’s brain is warped so that it fits with a standard template (the Montreal Neurological Institute or MNI template is most popular.)

Other techniques are also sometimes used, depending on the user’s preference and the software package.

Then the real fun begins: the stats. By far the most common statistical approach for detecting task-related neural activation is that based upon the General Linear Model (GLM), though there are alternatives.

We first need to define a model of what responses we’re looking for, which makes predictions as to what the neural signal should look like. The simplest model would be that the brain is more active at certain times, say, when a picture is on the screen. So our model would be simply a record of when the stimulus was on the screen. This is called a "boxcar" function (guess why):
In fact, we know that the neural response has a certain time lag. So we can improve our model by adding the canonical (meaning “standard”) haemodynamic response function (HRF).
Now consider a single voxel. The MRI signal in this voxel (the brightness) varies over time. If there were no particular neural activation in this area, we’d expect the variation to be purely noise:Now suppose that this voxel was responding to a stimulus present from time-point 40 to 80.
While the signal is on average higher during this period of activation, there’s still a lot of noise, so the data doesn’t fit with the model exactly.
The GLM is a way of asking, for each voxel, how closely it fits a particular model. It estimates a parameter, β, representing the “goodness-of-fit” of the model at that voxel, relative to noise. Higher β, better fit. Note that a model could be more complex than the one above. For example, we could have two kinds of pictures, Faces and Houses, presented on the screen at different times:
In this case, we are estimating two β scores for each voxel, β-faces and β-houses. Each stimulus type is called an explanatory variable (EV). But how do we decide which β scores are high enough to qualify as “activations”? Just by chance, some voxels which contain pure noise will have quite high β scores (even a stopped clock’s right twice per day!)

The answer is to calculate the t score, which for each voxel is β / standard deviation of β across the whole brain. The higher the t score, the more unlikely it is that the model would fit that well by chance alone. It’s conventional to finally convert the t score into the closely-related z score.

We therefore end up with a map of the brain in terms of z. z is a statistical parameter, so fMRI analysis is a form of statistical parametric mapping (even if you don’t use the "SPM" software!) Higher z scores mean more likely activation.

Note also that we are often interested in the difference or contrast between two EVs. For example, we might be interested in areas that respond to Faces more than Houses. In this case, rather than comparing β scores to zero, we compare them to each other – but we still end up with a z score. In fact, even an analysis with just one EV is still a contrast: it’s a contrast between the EV, and an “implicit baseline”, which is that nothing happens.

Now we still need to decide how high of a z score we consider “high enough”, in other words we need to set a threshold. We could use conventional criteria for significance: p less than 0.05. But there are 10,000 voxels in a typical fMRI scan, so that would leave us with 500 false positives.

We could go for a p value 10,000 times smaller, but that would be too conservative. Luckily, real brain activations tend to happen in clusters of connected voxels, especially when you’ve smoothed the data, and clusters are unlikely to occur due to chance. So the solution is to threshold clusters, not voxels.

A typical threshold would be “z greater than 2.3, p less than 0.05”, meaning that you're searching for clusters of voxels, all of which has a z score of at least 2.3, where there's only a 5% chance of finding a cluster that size by chance (based on this theory.) This is called a cluster corrected analysis. Not everyone uses cluster correction, but they should. This is what happens if you don't.

Thus, after all that, we hopefully get some nice colorful blobs for each subject, each blob representing a cluster and colour representing voxel z scores:

This is called a first-level, or single-subject, analysis. Comparing the activations across multiple subjects is called the second-level or group-level analysis, and it relies on similar principles to find clusters which significantly activate across most people.

This discussion has focused on the most common method of model-based detection of activations. There are other "data driven" or "model free" approaches, such as this. There are also ways of analyzing fMRI data to find connections and patterns rather than just activations. But that's another story...

What The Internet Thinks About Antidepressants

Toronto team Rizo et al offer a novel approach to psychopharmacology: trawling the internet for people's opinions. It's a rapid, web-based method for obtaining patient views on effects and side-effects of antidepressants.

They designed a script to Google the names of several antidepressants in the context of someone who's taking them, and checks to see if they describe any side-effects.

A large number of URLs were rapidly screened through Google Search™, using one server situated in Ohio, USA. The search strategy used language strings to denote active antidepressant drug usage, such as “I'm on [name of antidepressant]…” or “I
have been on [antidepressant] for ….”, or “I've started [antidepressant]…”, or “the [antidepressant] is giving me or causing me…”

They then used a thing called OpenCalais™ to read the search hits and decide whether they were mentioning particular diseases or symptoms. OpenCalais is a natural language processor which is meant to be able to automatically extract the meaning from text. However, to make sure it wasn't doing anything silly (natural language processing is quite tricky), they manually checked the results.

What happened? They found about 5,000 hits in total from people taking antidepressants, ranging from 210 for mirtazapine (Remeron) up to 835 for duloxetine (Cymbalta). That doesn't seem like all that many considering they searched on the entire internet, although they only searched English language websites.

Anyway, drowsiness, sleepiness or tiredness was mentioned in from 6.4% (duloxetine) down to 2.9% (fluoxetine) of the hits. Insomnia was noted in 4% (desvenlafaxine) down to 2.2% fluoxetine. And so on.

These results are a lot lower than anything previously reported from clinical trials, where the prevalence of drowsiness, for example, is often around 25% (vs. 10% on placebo); with some drugs, it's higher. So there's a big discrepancy, and it's hard to interpret these results. Maybe lots of people are having side effects and just not bothering to write about them. Or they're too embarrassed. Etc.

Still, it's a very clever idea it would probably be better used trying to discover which drugs work best. Neuroskeptic readers will know that clinical trials of antidepressants are flawed in several ways. I'd say they're actually better at telling us about side effects (which are probably roughly the same in clinical trials and in real life) than they are at telling us about efficacy (where this assumption doesn't hold)...

Links: There are many websites where people describe their experiences of medical treatments ranging from the fancy to the crude (but much more informative)...

Rizo C, Deshpande A, Ing A, & Seeman N (2010). A rapid, Web-based method for obtaining patient views on effects and side-effects of antidepressants. Journal of affective disorders PMID: 20705344

Is Your Brain Autistic?

There's been a lot of buzz and some scepticism about the

New brain scan to diagnose autism

Here's a quick overview. Autism is believed to be a disorder of brain development. If so, it should be possible to diagnose it based on a brain scan. Unfortunately, it's not. You can't tell, from a scan, whether someone has autism or not. Not even if you're a world expert.

There are reports of various differences between autistic and non-autistic brains - a bit smaller here, a bit bigger there - but there's a lot of overlap. So at present, diagnosis of autism is purely based on symptoms.

Ecker et al, a team based at the Institute of Psychiatry in South London, made use of a mathematical technique called a Support Vector Machine (SVM) to try to spot differences that the naked eye can't. An SVM is a learning algorithm: you "teach" it to spot differences by showing it lots of examples. In this case, they showed it 20 autistic brains, and the brains of 20 healthy controls matched for age, gender, and IQ.

How does an SVM work? Imagine that there are two kinds of, say, fruit. Both are kind of round but A's are more spherical than B's. So you could draw a plot of sphere-ness, and find a line separating A and B:
An SVM is an automatic method of finding that line. How? It's complex, but fortunately you don't need to know (I don't). Of course, that's easy, but imagine that things got more tricky. As well as the variable of roundness, there's colour. Fruit B can be either spherical and dark, or non-spherical and light (maybe it's two different stages of ripeness).

An SVM could do this easily too:
Now suppose that there's 1000 different variables, and you want to find the "line" - actually a 1000-dimensional "hyperplane" (a line is 2D, a plane is 3D, anything with 4D or more is a hyper-plane) - dividing the "space" of possibilities into two.

For a human that's impossible, but not for an SVM. This is essentially what Ecker et al did. Each dimension of their "space" was the amount of grey matter at a particular point in the brain. So, they were training the SVM to distinguish between autistic brains and non-autistic brains, based on their shape, but in a much more complex way than a human could.

Did it work? Surprisingly well. Here's the end result (the multi-dimensional space has been helpfully compressed into 2D by the SVM):
It wasn't perfect, but the best approach, based on the cortical thickness in the left hemisphere, managed 90% accuracy, which is pretty awesome. Focussing on the headline 90% result is cherry-picking a bit, because using other variables, like cortical curvature, wasn't as good, but even the worst ones managed 70-85%, much better than chance (50%). Importantly, they also tried the system on 20 adults with ADHD, and it classified them as non-autistic. This shows that it's not just measuring "normality".

*

Now the question everyone's asking: is this going to be used for diagnosis in the real world any time soon? The first thing to remember is that this is a scientific paper, and this result is first and foremost of research interest: it provides clues towards the biology, and ultimately the causes, of autism.

But let's suppose you're a clinician and you have someone who you suspect may have autism, but you're not sure. They're a tricky one, a borderline case. You use this system on their brain and it says they are autistic. Should that factor into your decision? It depends. The fact is that rather than an either-or result, the SVM returns a distance from the hyperplane for each brain. You can see this clearly in the plot above.

In my opinion, if you have a borderline case, and the machine says he's borderline, then that's not much help, and it doesn't matter if he's just over the line, or not quite over it. You already knew he was borderline.

But if the machine says that he's deep into the autism space, then I think that is something. It tells you that his brain is very typical of people with autism. Interestingly, Ecker et al found that distance from the hyperplane correlated with symptom severity for "social" and "communication" symptoms (though not "repetitive behaviours"). That's a pretty cool result because the SVM wasn't trained to do that, it was trained to decide on an either-or basis.

What needs to happen next? As it stands, this system only works for adults: it would fail for children or teenagers, because their brains are a very different size and shape. Exactly the same SVM approach could be used in younger age groups, though, so long as the patients and the controls were the same age.

We also need to make sure that the SVM can tell the difference between autism and other conditions; Ecker et al showed that it could distinguish autism from ADHD, but that's only one comparison and it might not be the hardest one: I would want to see it tested against things like epilepsy, mental retardation, and dyslexia as well.

Overall though, this is very exciting work, and certainly a cut above most "Brain Scans To Diagnose Mental Illness" studies that make it into the headlines.

Full Discloser: I know some of the researchers involved in this work.

Links: The same team had a paper out a few months back, using a slightly less sophisticated SVM approach, which managed 80% accuracy. I wrote about another application of SVMs previously: How To Read Minds. This study has been blogged about at The New Republic and Dormivigilia.

Ecker C, Marquand A, Mourão-Miranda J, Johnston P, Daly EM, Brammer MJ, Maltezos S, Murphy CM, Robertson D, Williams SC, & Murphy DG (2010). Describing the brain in autism in five dimensions--magnetic resonance imaging-assisted diagnosis of autism spectrum disorder using a multiparameter classification approach. Journal of Neuroscience, 30 (32), 10612-23 PMID: 20702694

Drugs for Starcraft Addiction

Are you addicted to Starcraft? Do you want to get off Battle.net and on a psychoactive drug?

Well, South Korean psychiatrists Han et al report that Bupropion sustained release treatment decreases craving for video games and cue-induced brain activity in patients with Internet video game addiction.

They took 11 people with "Internet Game Addiction" - the game being Starcraft, this being South Korea - and gave them the drug bupropion (Wellbutrin), an antidepressant that's also used in drug addiction and smoking cessation. These guys (because, predictably, they were all guys) were seriously hooked, playing on average at least 4 hours per day.

Six were absent from school because of playing Internet video game in Internet cafes for more than 2 months. Two IAGs had been divorced because of excessive Internet use at night.

They helpfully summarize Starcraft for the layperson:

As a military leader for one of three species, players must gather resources for training and expanding their species’ forces. Utilizing various strategies and alliances with other species, players attempt to lead their own species to victory.

Which is all true, but it doesn't quite communicate the sheer obsessiveness that's require to win this game. As Penny Arcade said "it is OCD masquerading as recreation", and that's coming from someone who literally plays video games for a living.

Anyway, apparently the drug worked:

After 6 weeks of bupropion SR treatment in the IAG group, there were significant decreases in terms of craving for playing StarCraft (23.6%), total playing game time (35.4%), and Internet Addiction Scale scores (15.4%)

They also did some fMRI and found that the addict's brains responded more strongly to pictures of Zerglings than did control people, and that the drug reduced activity a bit. But there was no placebo group, so we have no idea whether this was the drug or not.

Sadly, the point is moot, because Starcraft II has just come out, and it's more addictive than ever. I'm off to try and optimize my Terran build order, and by God I will get those 10 marines out in the first 5 minutes if it takes me all night...

Han DH, Hwang JW, & Renshaw PF (2010). Bupropion sustained release treatment decreases craving for video games and cue-induced brain activity in patients with Internet video game addiction. Experimental and clinical psychopharmacology, 18 (4), 297-304 PMID: 20695685

Very Severely Stupid About Depression

An unassuming little paper in the latest Journal of Affective Disorders may change everything in the debate over antidepressants: Not as golden as standards should be: Interpretation of the Hamilton Rating Scale for Depression.

Bear with me and I'll explain. It's less boring than it looks, trust me.

The Hamilton Scale (HAMD) is the most common system for rating the severity of depression. If you're only a bit down you get a low score, if you're extremely ill you get a high one. The maximum score's 52 but in practice it's extremely rare for someone to score more than 30.

First published in 1960, the HAMD is used in most depression research including almost all clinical trials of antidepressants. It's come under much criticism recently, but that's not the point here. The authors of the new paper, Kristen & von Wolff, simply asked: what does a given HAMD score mean in terms of severity?

It turns out that people have proposed no less than 5 different systems for interpreting HAMD scores. Do they all agree? Ha. Guess.

The pretty colors are mine. Just a glance shows a lot of variability, but the obvious outlier is the second one. That's the American Psychiatric Association (APA)'s official 2000 recommendations. Their interpretations of a given point on the scale tend to be worse than everyone else's.

This is most apparent at the top end. The APA use the terminology "Very Severe", which doesn't even appear on other scales. Much of what they class as "Very Severe" (23-26), two other scales class as "Moderate" depression! Amusingly, British authorities NICE seem to have been so unimpressed with this that they simply copied the APA's scale and toned everything down a notch for their 2009 criteria.

*

Why does this purely terminological debate matter? Well. A number of recent studies, most notoriously Kirsch et al (2008), have shown that antidepressants work better in more severe cases. See also my post here. The cut-off for antidepressants being substantially better than placebo generally comes out as about 26 on the HAMD in these studies.

Under the APA's 2000 terminology, this is well into the "Very Severe" band. Hence why Kirsch et al wrote - in a phrase that launched a thousand "Prozac Doesn't Work" headlines -

antidepressants reach... conventional criteria for clinical significance only for patients at the upper end of the very severely depressed category.

But for Bech, 26 is simply middle-of-the-road "major depression". For Furukawa, it's borderline "moderate" or "severe". Hmm. So if they'd gone with those criteria, Kirsch et al would have written instead

antidepressants reach... conventional criteria for clinical significance only for patients with major depression, of moderate-to-severe severity.

All of these terminological criteria are arbitrary, so this isn't necessarily more accurate, but it's no less so. The irony of the fact that Kirsch et al used the American Psychiatric Associations own criteria to skewer modern psychiatry isn't lost on me and probably wasn't lost on them either.

*

But where did the APA get their system from? This is the most extraordinary thing. Here's the paper they based their approach on. It's an 1982 British study by Kearns et al. The authors wanted to see how the HAMD compared to other depression scales. So they used lots of scales on the same bunch of depressed patients and compared them to each other, and to their own judgments of severity. Here's what they found:

You'll recognize the APA's categories, kind of, but they're all shifted. Why? We can only guess. Here's my guess. The scores in that Kearns et al graph were the average HAMD scores of people who fell into each severity band. The APA must have decided that they could use these to create cutoffs for severity.

How? It's not at all clear. The mean score for "Moderate" was 18, but that's the top end of Moderate in the APA's book; ditto for "Mild". The average "Very Severe" was 30 and the average "Severe" was 21 so the cut-off should have been 25 or 26 if you just went for the midpoint, in fact the APA went with 23. And so on.

That's before we get into the question of whether you should be using these results to make cutoffs at all (you shouldn't.) And the APA seem to have ignored the fact that the HAMD did not statistically significantly distinguish between "Severe" and "Moderate" depression anyway (p=0.1). Kearns et al's graph shows that other scales, like the Melancholia Subscale ("MS"), would be better. But everyone's been using the HAMD for the past 50 years regardless.

In Summary: Interpreting the Hamilton Scale is a minefield of controversy and the HAMD is far from a perfect scale of depression. Yet almost everything we know about depression and its treatment relies on the HAMD. Don't believe everything you read.

Kriston, L., & von Wolff, A. (2010). Not as golden as standards should be: Interpretation of the Hamilton Rating Scale for Depression Journal of Affective Disorders DOI: 10.1016/j.jad.2010.07.011

Kearns, N., Cruickshank, C., McGuigan, K., Riley, S., Shaw, S., & Snaith, R. (1982). A comparison of depression rating scales The British Journal of Psychiatry, 141 (1), 45-49 DOI: 10.1192/bjp.141.1.45

Hauser Of Cards

Update: Lots of stuff has happened since I wrote this post: see here for more.

A major scandal looks to be in progress involving Harvard Professor Marc Hauser, a psychologist and popular author whose research on the minds of chimpanzees and other primates is well-known and highly respected. The Boston Globe has the scoop and it's well worth a read (though you should avoid reading the comments if you react badly to stupid.)

Hauser's built his career on detailed studies of the cognitive abilities of non-human primates. He's generally argued that our closest relatives are smarter than people had previously believed, with major implications for evolutionary psychology. Now one of his papers has been retracted, another has been "corrected" and a third is under scrutiny. Hauser has also announced that he's taking a year off from his position at Harvard.

It's not clear what exactly is going on, but the problems seem to centre around videotapes of the monkeys that took part in Hauser's experiments. The story begins with a 2007 paper published in Proceedings of the Royal Society B. That paper has just been amended in a statement that appeared in the same journal last month:

In the original study by Hauser et al., we reported videotaped experiments on action perception with free ranging rhesus macaques living on the island of Cayo Santiago, Puerto Rico. It has been discovered that the video records and field notes collected by the researcher who performed the experiments (D. Glynn) are incomplete for two of the conditions.

The authors of the original paper were Hauser, David Glynn and Justin Wood. In the amendment, which is authored by Hauser and Wood i.e. not Glynn, they say that upon discovering the issues with Glynn's data, they went back to Puerto Rico, did the studies again, and confirmed that the original results were valid. Glynn left academia in 2007, to work for a Boston company, Innerscope Research, according to this online resume.

If that was the whole of the scandal it wouldn't be such a big deal, but according to the Boston Globe, that was just the start. David Glynn was also an author on a second paper which is now under scrutiny. It was published in Science 2007, with the authors listed as Wood, Glynn, Brenda Phillips and Hauser.

However, crucially, Glynn was not an author on the only paper which has actually been retracted, "Rule learning by cotton-top tamarins". This appeared in the journal Cognition in 2002. The three authors were Hauser, Daniel Weiss and Gary Marcus. David Glynn wasn't mentioned in the acknowledgements section either, and according to his resume, he didn't arrive in Hauser's lab until 2005.

So the problem, whatever it is, is not limited to Glynn.

Not was Glynn an author on the final paper mentioned in the Boston Globe, a 1995 article by Hauser, Kralik, Botto-Mahan, Garrett, and Oser. Note that the Globe doesn't say that this paper is formally under investigation, but rather, that it was mentioned in an interview by researcher Gordon G. Gallup who says that when he viewed the videotapes of the monkeys from that study, he didn't observe the behaviours which Hauser et al. said were present. Gallup is famous for his paper "Does Semen Have Antidepressant Properties?" in which he examined the question of whether semen... oh, guess.

The crucial issue for scientists is whether the problems are limited to the three papers that have so far been officially investigated or whether it goes further: that's an entirely open question right now.

In Summary: We don't know what is going on here and it would be premature to jump to conclusions. However, the only author who appears on all of the papers known to be under scrutiny, is Marc Hauser himself.

Hauser MD, Weiss D, & Marcus G (2002). Rule learning by cotton-top tamarins. Cognition, 86 (1) PMID: 12208654

Hauser MD, Glynn D, & Wood J (2007). Rhesus monkeys correctly read the goal-relevant gestures of a human agent. Proceedings. Biological sciences / The Royal Society, 274 (1620), 1913-8 PMID: 17540661

Wood JN, Glynn DD, Phillips BC, & Hauser MD (2007). The perception of rational, goal-directed action in nonhuman primates. Science (New York, N.Y.), 317 (5843), 1402-5 PMID: 17823353

Hauser MD, Kralik J, Botto-Mahan C, Garrett M, & Oser J (1995). Self-recognition in primates: phylogeny and the salience of species-typical features. Proceedings of the National Academy of Sciences of the United States of America, 92 (23), 10811-14 PMID: 7479889

A Time to Cry, and a Time to Laugh

This was trending on Twitter last night:

I feel really groggy and tired in the middle afternoon, but awake and energetic late at night. #idothistoo

I don't do Twitter but, ugh, fine, #idothistoo. However, in my case, the effect is sometimes more dramatic. If I'm in a depressive episode, my mood follows the same cycle, worse in the afternoon and better later in the evening, often to the point that some symptoms entirely disappear at nighttime.

In medical terms, this is called diurnal mood variation and it's considered a hallmark of clinical depression. The classical diurnal variation is progressive improvement throughout the day; waking up is said to be worst, especially when you wake up in the early hours of the morning (so-called "late insomnia").

In my experience, this is true but only when my depression is severe: I wake up two or three hours early feeling terrible, and then gradually improve. In milder episodes, I wake up at a normal time, or later than normal, and my mood is worse in the afternoon than the morning before recovering again.

Yet another phenomenon is the antidepressant effect of sleep deprivation. Staying awake the whole night often produces dramatic improvements in mood, though unfortunately the effect is transient and is lost when you do eventually fall asleep. This is unsurprising, if you think about classical diurnal mood variation: it's almost as if mood improves in proportion to the length of time spent awake. Again, I can confirm this from my personal experience.

Why does all this happen? No-one knows; many neurotransmitters and hormones have a circadian cycle - the best known being cortisol but almost everything is affected to some degree. Clearly a great many people experience diurnal cycles of energy - as Twitter shows - and the variations in depression are, presumably, an extreme form of the same phenomenon. The case of the man with almost no monoamines is also interesting: his symptoms showed a diurnal course, though it was reversed - better in the morning.

Diurnal variation is one of the few good things about depression. It's why the phrase "unrelenting misery" is not quite accurate: there is some relenting. You get to take a break, if only partial. It's even been suggested that it might be beneficial to schedule psychotherapy for the late evening, to maximize the mental energy available, and I can see how this would work, though it would rely on your therapist not having anything to better to do that night.

When depressed I've made use of this by staying up much later than usual; I generally go to bed around midnight but during an episode this often becomes more like 2 am, so as to squeeze as many hours of relative normality into the day.

Zapping Memory Better in Alzheimer's

Last month I wrote about how electrical stimulation of the hippocampus causes temporary amnesia - Zapping Memories Away.

Now Toronto neurologists Laxton et al have tried to use deep brain stimulation (DBS) to improve memory in people with Alzheimer's disease. Progressive loss of memory is the best-known symptom of this disorder, and while some drugs are available, they provide partial relief at best.

This study stems from a chance discovery by the same Toronto group. In 2008, they reported that stimulation of the hypothalamus caused vivid memory recollections a 50 year old man. In that case, the effect was entirely unintended and unexpected. The patient was being given DBS to try to curb his appetite (he weighed 420 pounds.) The hypothalamus is involved in regulating appetite, not memory - but the fornix, a nerve bundle that passes through that area, is. It's the main pathway connecting the hippocampus to the rest of the brain, and the hippocampus is vital for memory.

In this new study, Laxton et al implanted electrodes to stimulate the fornix in 6 patients with mild (early-stage) Alzheimer's. What happened? The results, unfortunately, were quite messy. On average, the patients symptoms got worse over the course of the year. Alzheimer's is a progressive degenerative disease, so this is what you'd expect to happen without treatment. The authors say that the decline was a bit slower than you'd expect in these kinds of patients, but to be honest, it's impossible to tell because there was no control group.

However, two patients did show memory improvements, and these were the same two who reported vivid recollections when the electrodes were first implanted (similar to the original obese guy):

Two of the 6 patients reported stimulation induced experiential phenomena. Patient 2 reported having the sensation of being in her garden, tending to the plants on a sunny day... Patient 4 reported having the memory of being fishing on a boat on a wavy blue colored lake with his sons and catching a large green and white fish. On later questioning in both patients, these events were autobiographical, had actually occurred in the past, and were accurately reported according to the patient’s spouse.

Also, the stimulation caused brain activation, generally switching "on" the areas that are turned "off" in Alzheimer's, and this lasted for a year (the length of the study so far). And there were no major side-effects. That's all good.

Overall, these results are extremely interesting, but we don't know how well the treatment really works, and we won't know until someone does a randomized controlled trial with a longer follow-up period; something which is, unfortunately, true of a lot of the latest DBS studies.

Link: The Neurocritic on the original 2008 paper.

Laxton AW, Tang-Wai DF, McAndrews MP, Zumsteg D, Wennberg R, Keren R, Wherrett J, Naglie G, Hamani C, Smith GS, & Lozano AM (2010). A phase I trial of deep brain stimulation of memory circuits in Alzheimer's disease. Annals of neurology PMID: 20687206