The Prefrontal Cortex Is Holistic

The question of whether the brain is "modular" - whether different parts do different things - has been a neuroscientific talking point since the days of the phrenologists.

They were the guys who believed that, not only were there modules, but that you could tell how big they were by measuring the shape of someone's skull, and so learn about their personality.

Phrenology made modules unfashionable for a while, but today they're back, and most of fMRI consists in trying to find areas of the brain that do different stuff, but in a new paper Wilson et al argue against taking modularism too far: Functional localization within the prefrontal cortex: missing the forest for the trees?

Their focus is the prefrontal cortex (PFC), a large chunk of the front of the brain which is bigger in humans than in any other species. The PFC is routinely subdivided into segments, each with (presumably) a different function. So we have the "emotional" vmPFC, the "memory" dlPFC, the "pleasure" OFC, etc.

Wilson et al don't dispute that there are some variations in function between different bits of the PFC, but they say that in all the excitement over localization, we may have overlooked the role of the PFC as a whole.

They discuss evidence from monkeys with PFC damage (or lesions which disconnect it from the rest of the brain). Damage to the entire PFC, they say, leaves monkeys completely unable to perform tasks which require storing concepts over time. For example, they can't learn that whenever they see, say, a red button, they ought to press it to get food. But if part of the PFC is intact, and it doesn't matter which part, monkeys can do this with only minor problems.

However, the PFC isn't required for all tasks. If the task only involves information which is all presented at once, the lesioned monkeys are OK. So they could learn, given a big panel covered in red buttons, to push the buttons to get food, because the buttons are all there simultaneously.

Hence the data from these tasks are congruent with the notion that [the PFC] is only crucial in memory during tasks requiring the processing of temporally complex events. This can be defined as an event to be learned about, in which information that is crucial to that learning is presented at more than one point in time, or that can only be interpreted with respect to a preceding event.

They say that evidence from human neuroimaging studies supports this view.

A meta-analysis has shown consistent recruitment of the same network of regions in the PFC across a range of cognitive demands. The authors argue that this supports specialization of function within the PFC, but of an unexpected nature, namely ‘a specific frontal-lobe network that is consistently recruited for solution of diverse cognitive problems’. The idea that large and different regions of the PFC are recruited by any task at hand supports our argument that the function of the PFC as a whole exceeds the sum of the functions of its subcomponents.

This all has echoes of Karl Lashley, an early neuroscientist (died 1958) who proposed the theory of "mass action" - that the whole cortex contributes to behaviour, rather than each part doing different things ("modularism").

Jerry Fodor, whose classic book The Modularity of Mind (1983) helped to rehabilitate modularism from its reputation as "phrenological", was also an advocate of this view - within limits.

Fodor argued that some brain systems, like vision, hearing and language, were cortical modules, but that above this, there was a non-modular system which was the basis for thought, intelligence and decision making. If I remember correctly, he didn't explicitly say that the prefrontal cortex was this system, but I'm sure he'd have no objections to Wilson et al's account.

Wilson CR, Gaffan D, Browning PG, & Baxter MG (2010). Functional localization within the prefrontal cortex: missing the forest for the trees? Trends in neurosciences PMID: 20864190

Big Pharma Explain How To Pick Cherries

Here at Neuroskeptic, we see a lot of bad science. Maybe, over the years (all 2 of them) that I've been writing this blog, I've become a bit jaded. Maybe I'm less distressed by it than I used to be. Cynical, even.

But this one really takes the biscuit. And then it takes the tin. And relieves itself in it: A New Population-Enrichment Strategy to Improve Efficiency of Placebo-Controlled Clinical Trials of Antidepressant Drugs.

Don't worry - it's from a big pharmaceutical company (GlaxoSmithKline), so I don't have to worry about hurting feelings.

It's is full to bursting with colourful graphs and pictures, but the basic idea is very simple. As in "simpleton".

Suppose you're testing a new drug against placebo. You decide to do a multicentre trial, i.e. you enlist lots of doctors to give the drug, or placebo, to their patients. Each clinic or hospital which takes part is a "centre". Multicentre trials are popular because they're an easy way of quickly testing a drug on a large number of patients.

Anyway, suppose that the results come in, and it turns out that the drug didn't work any better than placebo, which unfortunately is what happens rather often in modern trials of antidepressants. Oh dear. The drug's crap. That's the end of that chapter.

...or is it?!? say GSK. Maybe not. They have a clever trick. Look at the results from each centre individually. Placebo response rates will probably vary between centres: in some of them, the placebo people don't get better, in others, they get lots better.

Now, suppose that you just chucked out all of the data from centres where the people on placebo got much better, on the grounds that there must be something weird going on in those ones. They reanalyzed the data from 1,837 patients given paroxetine or placebo, across 124 centres. In the dataset as a whole, paroxetine barely outperformed placebo. However, in the centres where people on placebo only improved a little, the drug was much better than placebo!

Well, of course it was. Imagine that the drug has no effect. Some people just get better and others don't. Let's assume that each person randomly gets between 0 and 25 better, with an equal chance of any outcome. Half are on drug and half are on placebo, but it makes no difference.

Let's further assume that there are 50 centres, with 20 people per centre (1000 people total). I knocked up a "simulation" of this in Excel (it took 10 minutes). Here's what you get:

The blue dots show, for each imaginary centre, drug improvement vs. placebo improvement. There's no correlation (it's random), and, on average, there is no difference: both average out at 12 points. The drug doesn't work.

The red dots show the "Treatment Effect" i.e. [drug improvement - placebo improvement]. The average is 0 - because the drug doesn't work. But there's a strong negative correlation between Treatment Effect and the placebo improvement - in centres where people improved lots on placebo, the drug worked worse.

This is exactly what Glaxo show in Figure 1a (see above). They write:

The analysis of the surface response indicated the predominant role of center specific placebo response as compared with the dose strength in determining the Treatment Effect of paroxetine.

But of course they correlate. You're correlating placebo improvement with itself: the "Treatment Effect" is a function of the placebo improvement. It's classic regression to the mean.

Of course if you chuck out the centres where people on placebo do well (the grey box in my picture), the drug seems to work pretty nicely. But this is cheating. It is cherry-picking. It is completely unscientific. (To give the authors their due, they also eliminated the centres where the placebo response was very low. This could, under some assumptions, make the analysis unbiased, but they don't show that this was their intention, let alone that it would eliminate all of the bias.)

The authors note that this could be a source of bias, but say that it wouldn't be one if it was planned out in advance: "in order to overcome the bias risk, the enrichment strategy should be accounted for and pre-planned in the study protocol." This is like saying that if you announce, before playing chess, that you are going to cheat, it's not cheating.

To be fair to the authors, assuming the drug does work, this method would improve your chances of correctly detecting the effect. Centres with very high placebo responses quite possibly are junk. Assuming the drug works.

But if we're assuming the drug works, why are we bothering to do a trial? The whole point of a trial is to discover something we don't know. The authors justify their approach by suggesting that it would be useful for drug companies who want to do a "proof-of-concept" trial to find out whether an experimental drug might work under the most favourable conditions, i.e. whether they should bother continuing to research it.

They say that such trials "are inherently exploratory in their conception, aimed at signal detection, open to innovation..." - in other words, that they're not meant to be as rigorous as late-stage trials.

Fair enough. But this method is not even suitable for proof-of-concept, because it would (as I have shown above in my 10 minute simulation) increase your chance of finding an "effect" from a drug that doesn't work.

Whatever the truth is, this method will give the same result, so it's not useful evidence. It's like saying "Heads I win, tails you lose". You've set it up so that I lose - the coin toss doesn't tell us anything.

All of the author's results are based on trials in which the drug "should have worked": they do not appear to have simulated what would happen if they used this method on trials where it didn't work, as I just did. So I'm doing Pharma a big favour by writing this post, because if they adopt this approach, they're more likely to waste money on drugs that don't work.

They should be paying me for this stuff.

Merlo-Pich E, Alexander RC, Fava M, & Gomeni R (2010). A New Population-Enrichment Strategy to Improve Efficiency of Placebo-Controlled Clinical Trials of Antidepressant Drugs. Clinical Pharmacology and Therapeutics PMID: 20861834

Sociopathic Dementia

Frontotemporal dementia (FTD) is a tragic, but scientifically fascinating, disease.


FTD only accounts for a small fraction of dementias in total (estimates range from 2% to 10%), but it typically strikes people aged in their 50s or 60s, i.e. much earlier than the average for Alzheimer's disease, the most common cause of dementia. As a result, FTD accounts for a large proportion of early-onset cases.

The symptoms are different to those of Alzheimer's, at least in the early stages. Memory problems and confusion are not prominent. Nor are hallucinations and delusions, which are seen in 20% of Alzheimer's, but only 2% of FTD.

Instead, patients often present with language problems - either forgetting what words mean, starting with uncommon words and progressing to easy ones ("semantic dementia"), or losing the ability to articulate speech ("nonfluent aphasia").

But the most disturbing effects are behavioural and personality changes. These are not seen in all cases, but in some people (the "behavioural variant"), they are the main symptom. Patients may begin to act entirely out of character, including criminal acts.

Aggressive behaviour is also sometimes seen in Alzheimer's, but it's usually associated with confusion or hallucinations: people "don't know what they're doing". In FTD, patients can commit serious crimes even though their cognitive function is pretty much intact: they do know what they're doing.

Mario F. Mendez discusses this in a new paper, The Unique Predisposition to Criminal Violations in Frontotemporal Dementia, and asks whether people who commit crimes while suffering from FTD should be considered legally responsible for their apparantly "sociopathic" actions. He presents 4 case histories.

Patient 1: A left-handed male in his sixties began stalking and attempting to molest children for the first time in his life. He followed children home from school and tried to touch them... On another occasion, he stood at the foot of a pool and stared at the children for a prolonged time.

When he exposed himself to his neighbor’s children, he was arrested. The patient did not deny his actions, could describe them in detail, and endorsed them as wrong and harmful. Despite this, he stated that he did not feel that he was causing harm at the time of his acts.

The patient’s personality had deteriorated over the prior four years, with decreased concern for others, disinhibition, and compulsive hoarding. He had caused disturbances at work, such as intruding into others’ conversations and walking into others’ offices... constantly pilfering... hiding money.... In addition, he ate indiscriminately, even going through waste containers and eating garbage. He stopped showering and wore the same clothes every day.

Neuropsychological testing and brain scans suggested early FTD, and his mother had reportedly suffered unspecified dementia; FTD is often genetic. He was not prosecuted. This case has a lot in common with the man who became a pedophile after surgery for a brain tumour: not just the pedophilia, but other symptoms like compulsive hoarding, over-eating, etc.

Patient 4: A right-handed man in his early fifties had a hit-and-run accident and left the scene without concern. He had struck a van with passengers but kept driving. The police stopped him a short distance away from the scene, and he did not deny his action.

Leaving the scene of an accident was not characteristic of his premorbid personality, yet he had had several recent traffic violations... He could recall and describe the accident, knew that it was wrong to leave the scene, but did not feel the need to stop at the time.

Over the prior two years, the patient’s pervasive behavior had significantly changed. He had become disengaged and emotionally detached; for example, he did not react to the death of his mother...

He was no longer embarrassed over passing gas or belching in public or appearing partially clothed in front of others. The patient had a tendency toward hyperorality, especially for peanuts, and had a decline in personal hygiene. Other aspects of the history included dysarthria and a recent tendency to choke on liquids.

This patient showed clear signs of motor neuron disease, which occurs in up to 15% of FTD cases. He died, as a result of the progression of the motor neuron disease, one year later, after developing other symptoms of FTD. His death meant he could not be tried for the hit-and-run.

Mendez notes that legally, these patients would probably not qualify for the "insanity defence". Under the British M'Naghten Rules, also adopted by the USA, the defendant is only eligible if they were

labouring under such a defect of reason, from disease of the mind, as not to know the nature and quality of the act he was doing; or, if he did know it, that he did not know he was doing what was wrong.

These patients do not fit that bill.

Finally, why does FTD cause sociopathic behaviour? Mendez says that it is because it involves degeneration of the vmPFC, linking FTD patients to the classic case of Phineas Gage whose vmPFC was destroyed by a flying iron rod. But Gage, while he did show personality changes, actually managed to function fairly well in society.

So temporal lobe degeneration probably also contributes to the FTD behavioural syndrome, especially since many of the symptoms (like compulsive eating) are seen in monkeys with temporal lobe lesions.

Mendez MF (2010). The unique predisposition to criminal violations in frontotemporal dementia. The journal of the American Academy of Psychiatry and the Law, 38 (3), 318-23 PMID: 20852216

The Rise of the Mouse

Everyone knows that scientists experiment on rats, and guinea-pigs. That's why we have "lab rats" and why, if you're trying out something new, you're a "human guinea-pig".

But this is all out of date. Nowadays, mice are the most popular lab animals. Here's a graph showing the number of scientific papers published each year, mentioning each kind of critter (data gathered with this script):

Rats were on top until about 10 years ago, when mice overtook them. Why? No-one wants to study mice if they can help it: they are horrible to work with compared to rats, and rats are more similar to humans physiologically. This is why rats were more popular for a long time. (Contrary to popular belief, guinea pigs were never used all that much, and they've become even less popular with the rise of mice.)

Non-scientists tend to think of rats as just big mice. They're not: mice are less intelligent, harder to handle (they bite... a lot), and they smell bad. The fact that they're smaller makes surgery, and even simple stuff like taking blood samples, much harder. On the plus side, you can fit more of them in any given space, making them cheaper, but that's about it.

So why did mice suddenly claim the crown? One word - knockout. Mice are the only mammal in which it's easy to perform genetic knockout, i.e. eliminating the function of a single gene. It's extremely difficult in rats, because, for reasons no-one really understands, it is harder to get rat stem cells to grow in vitro.

Knockout mice were "invented" in 1989, and the inexorable rise in the number of mouse papers began a few years later. Recently, there have been reports that knockout rats may now be easy; whether this will lead to a rat renaissance remains to be seen.

Knockouts have revolutionized biology, because they make it easy to investigate what each gene does. Just knock it out, and see what's wrong with your mouse. This is why there are mouse models of so many genetic diseases, while rat and monkey models are only available for a few disorders.

The Refrigerator Mother

Autism is biological: that's the one thing everyone agrees about it. Scientific orthodoxy is that it's a neurodevelopmental condition caused by genetics, in most cases, and by environmental insult, such fetal exposure to anticonvulsants, in rare cases. Jenny McCarthy orthodoxy is that "toxins" - usually in vaccines - are to blame, not genes, and that the underlying damage might be in the gut not the brain: but they agree that it's biological.

However, it hasn't always been this way. From the 1950s to about the 1980s, there was a widespread view that autism was a purely psychological condition. Bruno Bettelheim is the name most often linked to this view. Bettelheim spent most of his career at the University of Chicago's Orthogenic School, an institution for "disturbed" children, including autistics as well as "schizophrenic" and others.

His magnum opus was his book The Empty Fortress: Infantile Autism and the Birth of the Self, in which he outlined his theory of autism illustrated by three long case histories. His ideas are now referred to as the "refrigerator mother" theory.

For Bettelheim, autism was a reaction to severe neglect. Not of physical needs, which would be fatal, but of emotional relations. In his view, the most common underlying cause of this neglect was when the mother (and to a lesser extent, the father) did not want the child to exist. They cared for him, but they did so in a mechanical fashion, treating the baby as a mouth to feed and a nappy to change, rather than as a human being.

Hence the "refrigerator" - it provides food, but it's cold.

The result was that the child never learned to interact with the mother on anything other than a mechanical level; and for Bettelheim, as for most psychoanalysts, our relationships with our parents were the model on which all our other relationships were based.

The mechanical mother thus left the autistic child unable to relate to anyone, indeed, unable to conceive of the existence of other human beings, and thus lacking a sense of "self" as opposed to "others".


The repetitive behaviours and obsessive interests characteristic of autism were seen as an active, even heroic, coping strategy. They were the child's way of asserting what little self they had, by doing something for themselves, albeit something "pointless". But they also had symbolic meanings: "Joey's" interest in fans, propellers and other rotating objects was interpreted as a representation of the "vicious circle" of his life. And so on.

*

Bettelheim's ideas are now generally derided as dangerously wrong; his reputation suffered a hit when, after his suicide in 1990, stories emerged from former colleagues and patients painting him in a nasty light. But psychiatry's wider turn away from Freud and towards biology probably made his downfall inevitable.

Today the "refrigerator mother theory" is routinely cited as a cautionary tale of how deeply one can misunderstand autism. Ironically, Bettelheim's only reference to that term in The Empty Fortress is a quotation, from none other than Leo Kanner, the man who coined the term 'childhood autism' in 1944. Kanner referred to the "emotional refrigeration" he observed in the families of autistic children, although it's not clear that he thought of it as causing the autism.

There is no doubt that Bettelheim's approach was unscientific. He repeatedly claimed that the fact that many children improved after three or four years at the Orthogenic School proved that their autism was psychological, because if it were biological it would be permanent.

Yet there is no reason to assume that children with a neurodevelopmental disorder would never change as they grew up. There was no control group, let alone a placebo group, to show that the children wouldn't have "grown out of" some symptoms anyway. (Edit: In fact, Kanner himself had written about improvement with age way back in 1943, in the first ever paper about autistic children! So there was simply no excuse for Bettelheim's flawed argument.)

Bettelheim's attributing the cause of autism to family dynamics was post hoc: for each autistic child, he looked back into their family history (i.e. what the parents reported) and found that they "consciously or unconsciously" didn't want the child to exist.

Yet all this proves is that it is possible to interpret a parent's behaviour in that way, in retrospect, if you want to. The "or unconsciously" caveat creates endless scope for over-interpretation.

But even if we now see autism as a neurodevelopmental disorder, there is something attractive about Bettelheim's book: it seems to be a serious attempt to understand the autistic experience "from the inside", and to appreciate the autistic child as a person rather than a disease. This is something that we rarely see nowadays.

Bettelheim's problem was that he tried to understand autistic behaviour from the assumption that the autistic child was, deep down, entirely "normal". Hence his interpretation of, say, Joey's fascination with rotating objects as symbolic of his life situation (and also as reflecting the fact that his father was often flying away in propeller-driven aircraft, which he was).

Yet couldn't it be that Joey was just fascinated by spinning fans per se? There's nothing interesting about rotating objects. They must have a hidden meaning. Otherwise it makes no sense - to someone who isn't autistic. But all that means is that trying to understand the autistic child is rather difficult if you don't bear in mind that they are autistic.

The Horror, The Horror (Movies)

Previously, I blogged about how the placebo effect is at work when you watch a horror movie. As part of my... research for this post I watched quite a lot of them. Here are my thoughts on some recent ones. Roughly in order of best to worst. Some minor spoilers, but nothing worse than you'd get from the trailer.

• The Last Exorcism - Very scary, and full of surprises. The trailer makes it look like a shameless clone of The Exorcist; it isn't. Gets Neuroskeptic bonus points for the opening bit where the hero, a preacher who's lost his faith, talks about how he keeps doing "exorcisms" because it's psychologically helpful i.e. a placebo effect.
• The Broken - Inventive, intelligent and creepy. A woman starts seeing her own doppelgänger after a car crash. Neuroskeptic bonus points for mentioning Capgras syndrome, a classic neurological disorder. Well worth watching.
• The Signal - A mysterious TV glitch sends people crazy. But who's actually affected, and who's just been sent crazy by the fact that everyone else is going crazy around them? A clever twist on the zombie apocalypse genre, and manages to be both frightening and funny.
  
• Carriers - An airborne Ebola virus wipes out almost everyone. Four teens try to escape. The characters and acting are pretty blah, but the concept is good, and it's well produced.
  
• Dread - Student film-makers decide to make a documentary about people's worst fears, but one of them is a psychopath, so they end up making a Saw movie. Good, if a bit predictable.
  
• The House of the Devil - An attempt at the kind of anticipation-horror that I talked about in my past post - nothing really happens, but the build-up is tense. Up to a point. Then it goes on for another half hour and gets tedious. Missable.
  
• Tell Tale - Pretty standard slasher, except that the serial killer... is an internal organ! Actually not all that bad, but nothing special.
  
• Mutants - Zombies attack survivors holed up in a hospital. In France. Extremely generic, there is no point in watching this if you've seen, well, any other zombie movie from the past 5 years.
• Mulberry Street - "They're rat people, they're f-king rat people!" Some virus strikes New York, turning people into rat people. Who are also psychopaths. More funny than scary, unintentionally. Apparently this was done on basically zero budget: what's disappointing is that it comes across as quite polished despite that - the budget isn't the problem, the script is.

A Tale of Two Genes

An unusually gripping genetics paper from Biological Psychiatry: Pagnamenta et al.

The authors discuss a family where two out of the three children were diagnosed with autism. In 2009, they detected a previously unknown copy number variant mutation in the two affected brothers: a 594 kb deletion knocking out two genes, called DOCK4 and IMMP2L.

Yet this mutation was also carried by their non-autistic mother and sister, suggesting that it wasn't responsible for the autism. The mother's side of the family, however, have a history of dyslexia or undiagnosed "reading difficulties"; all of the 8 relatives with the mutation "performed poorly on reading assessment".

Further investigation revealed that the affected boys also carried a second, entirely separate, novel deletion, affecting the gene CNTNAP5. Their mother and sister did not. This mutation came from their father, who was not diagnosed with autism but apparently had "various autistic traits".

Perhaps it was the combination of the two mutations that caused autism in the two affected boys. The mother's family had a mutation that caused dyslexia; the father's side had one that caused some symptoms of autism but was not, by itself, enough to cause the disorder per se.

However, things aren't so clear. There were cases of diagnosed autism spectrum disorders in the father's family, although few details are given and DNA was only available from one of the father's relatives. So it may have been that the autism was all about the CNTNAP5, and this mutation just has a variable penetrance, causing "full-blown" autism in some people and merely traits in others (like the father).

In order to try to confirm whether these two mutations do indeed cause dyslexia and autism, they searched for them in several hundred unrelated autism and dyslexia patients as well as healthy controls. They detected the a DOCK4 deletion in 1 out of 600 dyslexics (and in his dyslexic father, but not his unaffected sister), but not in 2000 controls. 3 different CNTNAP5 mutations were found in the affected kids from 3 out of 143 autism families, although one of them was also found in over 1000 controls.

This is how psychiatric genetics is shaping up: someone finds a rare mutation in one family, they follow it up, and it's only carried by one out of several hundred other cases. So there are almost certainly hundreds of genes "for" disorders like autism, and it only takes a mutation in one (or two) to cause autism.

Here's another recent example: they found PTCHD1 variants in a full 1% of autism cases. It seems to me that autism, for example, is one of the things that happens when something goes wrong during brain development. Hundreds of genes act in synchrony to build a brain; it only takes one playing out of tune to mess things up, and autism is one common result.

Mental retardation and epilepsy are the other main ones, and we know that there are dozens or hundreds of different forms of these conditions each caused by a different gene or genes. The million dollar question is what it is that makes the autistic brain autistic, as opposed to, say, epileptic.

The "rare variants" model has some interesting implications. The father in the Pagnamenta et al. study had never been diagnosed with anything. He had what the authors call "autistic traits", but presumably he and everyone just thought of those as part of who he was - and they could have been anything from shyness, to preferring routine over novelty, to being good at crosswords.

Had he not carried the CNTNAP5 mutation, he'd have been a completely different person. He might well have been drawn to a very different career, he'd probably never have married the woman he did, etc.

Of course, that doesn't mean that it's "the gene for being him"; all of his other 23,000 genes, and his environment, came together to make him who he was. But the point is that these differences don't just pile up on top of each other; they interact. One little change can change everything.

Link: BishopBlog on why behavioural genetics is more complicated than some people want you to think.

Pagnamenta, A., Bacchelli, E., de Jonge, M., Mirza, G., Scerri, T., Minopoli, F., Chiocchetti, A., Ludwig, K., Hoffmann, P., & Paracchini, S. (2010). Characterization of a Family with Rare Deletions in CNTNAP5 and DOCK4 Suggests Novel Risk Loci for Autism and Dyslexia Biological Psychiatry, 68 (4), 320-328 DOI: 10.1016/j.biopsych.2010.02.002

Stopping Antidepressants: Not So Fast

People who quit antidepressants slowly, by gradually decreasing the dose, are much less likely to suffer a relapse, according to Baldessarini et al. in the American Journal of Psychiatry.

They describe a large sample (400) of patients from Sardinia, Italy, who had responded well to antidepressants, and then stopped taking them. The antidepressants had been prescribed for either depression, or panic attacks.

People who quit suddenly (over 1-7 days) were more likely to relapse, and relapsed sooner, than the ones who stopped gradually (over a period of 2 weeks or more).

This graph shows what % of the patients in each group remained well at each time point (in terms of days since their final pill.) As you can see, the two lines separate early, and then remain apart by about the same distance (20%) for the whole 12 months.

What this means is that rapid discontinuation didn't just accelerate relapses that were "going to happen anyway". It actually caused more relapses - about 1 in 5 "extra" people. These "extra" relapses all happened in the first 3 months, because after that, the slope of the lines is identical.

On the other hand, they rarely happened immediately - it's not as if people relapsed within days of their last pill. The pattern was broadly similar for older antidepressants (tricyclics) and newer ones (SSRIs).

The authors note that these data throw up important questions about "relapse prevention" trials comparing people who stay on antidepressants vs. those who are switched - abruptly - to placebo. People who stay on the drug usually do better, but is this because the drug works, or because the people on placebo were withdrawn too fast?

This was an observational study, not an experiment. There was no randomization. People quit antidepressants for various "personal or clinical reasons"; 80% of the time it was their own decision, and only 20% of the time was it due to their doctor's advice.

So it's possible that there was some underlying difference between the two groups, that could explain the differences. Regression analysis revealed that the results weren't due to differences in dose, duration of treatment, diagnosis, age etc., but you can't measure every possible confound.

Only randomized controlled trials could provide a final answer, but there's little chance of anyone doing one. Drug companies are unlikely to fund a study about how to stop using their products. So we have only observational data to go on. These data fit in with previous studies showing that there's a similar story when it comes to quitting lithium and antipsychotics. Gradual is better.

But that's common sense. Tapering medications slowly is a good idea in general, because it gives your system more time to adapt. Of course, sometimes there are overriding medical reasons to quit quickly, but apart from in such cases, I'd always want to come off anything as gradually as possible.

Baldessarini RJ, Tondo L, Ghiani C, & Lepri B (2010). Illness risk following rapid versus gradual discontinuation of antidepressants. The American journal of psychiatry, 167 (8), 934-41 PMID: 20478876

Shotgun Psychiatry

There's a paradox at the heart of modern psychiatry, according to an important new paper by Dr Charles E. Dean, Psychopharmacology: A house divided.

It's a long and slightly rambling article, but Dean's central point is pretty simple. The medical/biological model of psychiatry assumes that there are such things as psychiatric diseases. Something biological goes wrong, presumably in the brain, and this causes certain symptoms. Different pathologies cause different symptoms - in other words, there is specificity in the relationship between brain dysfunction and mental illness.

Psychiatric diagnosis rests on this assumption. If and only if we can use a given patient's symptoms to infer what kind of underlying illness they have (schizophrenia, bipolar disorder, depression), diagnosis makes sense. This is why we have DSM-IV which consists of a long list of disorders, and the symptoms they cause. Soon we'll have DSM-V.

The medical model has been criticized and defended at great length, but Dean doesn't do either. He simply notes that modern psychiatry has in practice mostly abandoned the medical model, and the irony is, it's done this because of medicines.

If there are distinct psychiatric disorders, there ought to be drugs that treat them specifically. So if depression is a brain disease, say, and schizophrenia is another, there ought to be drugs that only work on depression, and have no effect on schizophrenia (or even make it worse.) And vice versa.

But, increasingly, psychiatric drugs are being prescribed for multiple different disorders. Antidepressants are used in depression, but also all kinds of anxiety disorders (panic, social anxiety, general anxiety), obsessive-compulsive disorder, PTSD, and more. Antipsychotics are also used in mania and hypomania, in kids with behaviour problems, and increasingly in depression, leading some to complain that the term "antipsychotics" is misleading. And so on.

So, Dean argues, in clinical practice, psychiatrists don't respect the medical model - yet that model is their theoretical justification for using psychiatric drugs in the first place.

He looks in detail at one particularly curious case: the use of atypical antipsychotics in depression. Atypicals, like quetiapine (Seroquel) and olanzapine (Zyprexa), were originally developed to treat schizophrenia and other psychotic states. They are reasonably effective, though most of them are no more so than older "typical" antipsychotics.

Recently, atypicals have become very popular for other indications, most of all mood disorders: mania and depression. Their use in mania is perhaps not so surprising, because severe mania has much in common with psychosis. Their use in depression, however, throws up many paradoxes (above and beyond how one drug could treat both mania and its exact opposite, depression.)

Antipsychotics block dopamine D2 receptors. Psychosis is generally considered to be a disorder of "too much dopamine", so that makes sense. The dopamine hypothesis of psychosis and antipsychotic action is 50 years old, and still the best explanation going.

But depression is widely considered to involve too little dopamine, and there is lots of evidence that almost all antidepressants (indirectly) increase dopamine release. Wouldn't that mean that antidepressants could cause psychosis (they don't?). And why, Dean asks, would atypicals, that block dopamine, help treat depression?

Maybe it's because they also act on other systems? On top of being D2 antagonists, atypicals are also serotonin 5HT2A/C receptor blockers. Long-term use of antidepressants reduces 5HT2 levels, and some antidepressants are also 5HT2 antagonists, so this fits. However, it creates a paradox for the many people who believe that 5HT2 antagonism is important for the antipsychotic effect of atypicals as well - if that were true, antidepressants should be antipsychotics as well (they're not.) And so on.

There may be perfectly sensible answers. Maybe atypicals treat depression by some mechanism that we don't understand yet, a mechanism which is not inconsistent with their also treating psychosis. The point is that there are many such questions standing in need of answers, yet psychopharmacologists almost never address them. Dean concludes:

it seems increasingly obvious that clinicians are actually operating from a dimensional paradigm, and not from the classic paradigm based on specificity of disease or drug... the disjunction between those paradigms and our approach to treatment needs to be recognized and investigated... Bench scientists need to be more familiar with current clinical studies, and stop using outmoded clinical research as a basis for drawing conclusions about the relevance of neurochemical processes to drug efficacy. Bench and clinical scientists need to fully address the question of whether the molecular/cellular/anatomical findings, even if interesting and novel, have anything to do with clinical outcome.

Dean CE (2010). Psychopharmacology: A house divided. Progress in neuro-psychopharmacology & biological psychiatry PMID: 20828593

You're (Brain Is) So Immature

How mature are you? Have you ever wanted to find out, with a 5 minute brain scan? Of course you have. And now you can, thanks to a new Science paper, Prediction of Individual Brain Maturity Using fMRI.

This is another clever application of the support vector machine (SVM) method, which I've written about previously, most recently regarding "the brain scan to diagnose autism". An SVM is a machine learning algorithm: give it a bunch of data, and it'll find patterns in it.

In this case, the input data was brain scans from children, teenagers and adults, and the corresponding ages of each brain. The pattern the SVM was asked to find was the relationship between age and some complex set of parameters about the brain.

The scan was resting state functional connectivity fMRI. This measures the degree to which different areas of the brain tend to activate or deactivate together while you're just lying there (hence "resting"). A high connectivity between two regions means that they're probably "talking to each other", although not necessarily directly.

It worked fairly well:

Out of 238 people aged 7 to 30, the SVM was able to "predict" age pretty nicely on the basis of the resting state scan. This graph shows chronological age against predicted brain age (or "fcMI" as they call it). The correlation is strong: r2=0.55.

The authors then tested it on two other large datasets: one was resting state, but conducted on a less powerful scanner (1.5T vs 3.0T) (n=195), and the other was not designed as a resting state scan at all, but did happen to include some resting state-like data (n=186). Despite the fact that these data were, therefore, very different to the original dataset, the SVM was able to predict age with r2 over 0.5 as well.

*

What use would this be? Well, good question. It would be all too easy to, say, find a scan of your colleague's brain, run it through the Mature-O-Meter, and announce with glee that they have a neurological age of 12, which explains a lot. For example.

However, while this would be funny, it wouldn't necessarily tell you anything about them. We already know everyone's neurological age. It's... their age. Your brain is an old as you are. These data raise the interesting possibility that people with a higher Maturity Index, for their age, are actually more "mature" people, whatever that means. But that might not be true at all. We'll have to wait and see.

How does this help us to understand the brain? An SVM is an incredibly powerful mathematical tool for detecting non-linear correlations in complex data. But just running an SVM on some data doesn't mean we've learned anything: only the SVM has. It's a machine learning algorithm, that's what it does. There's a risk that we'll get "science without understanding" as I've written a while back.

In fact the authors did make a start on this and the results were pretty neat. They found that as the brain matures, long-range functional connections within the brain become stronger, but short-range interactions between neighbours get weaker and this local disconnection with age is the most reliable change.

You can see this on the pic above: long connections get stronger (orange) while short ones get weaker (green), in general. This is true all across the brain.

It's like how when you're a kid, you play with the kids next door, but when you grow up you spend all your time on the internet talking to people thousands of miles away, and never speak to your neighbours. Kind of.

Link: Also blogged about here.

Dosenbach NU, Nardos B, Cohen AL, Fair DA, Power JD, Church JA, Nelson SM, Wig GS, Vogel AC, Lessov-Schlaggar CN, Barnes KA, Dubis JW, Feczko E, Coalson RS, Pruett JR Jr, Barch DM, Petersen SE, & Schlaggar BL (2010). Prediction of individual brain maturity using fMRI. Science (New York, N.Y.), 329 (5997), 1358-61 PMID: 20829489

"Koran Burning"

According to the BBC:

Koran protests sweep Afghanistan... Thousands of protesters have taken to the streets across Afghanistan... Three people were shot when a protest near a Nato base in the north-east of the country turned violent.

Wow. That's a lot of fuss about, literally, nothing - the Koran burning hasn't happened. So what are they angry about? The "Koran Burning" - the mere idea of it. That has happened, of course - it's been all over the news.

Why? Well, obviously, it's a big deal. People are getting shot protesting about it in Afghanistan. It's news, so of course the media want to talk about it. But all they're talking about is themselves: the news is that everyone is talking about the news which is that everyone is talking about...

A week ago no-one had heard of Pastor Jones. The only way he could become newsworthy is if he did something important. But what he was proposing to do was not, in itself, important: he was going to burn a Koran in front of a handful of like-minded people.

No-one would have cared about that, because the only people who'd have known about it would have been the participants. Muslims wouldn't have cared, because they would never have heard about it. "Someone You've Never Heard Of Does Something" - not much of a headline.

But as soon as it became news, it was news. Once he'd appeared on CNN, say, every other news outlet was naturally going to cover the story because by then people did care. If something's on CNN, it's news, by definition. Clever, eh?


What's odd is that Jones actually announced his plans way back in July; no-one took much notice at the time. Google Trends shows that interest began to build only in late August, peaking on August 22nd, but then falling off almost to zero.

What triggered the first peak? It seems to have been the decision of the local fire department to deny a permit for the holy book bonfire, on August 18th. (There were just 6 English-language news hits between the 1st and the 17th.)

It all kicked off when the Associated Press reported about the fire department's decision on August 18th and was quickly followed up by everyone else; the AP credit the story to the local paper The Gainsville Sun who covered the story on the same day.

But in their original article, the Sun wrote that Pastor Jones had already made "international headlines" over the event. Indeed there were a number of articles about it in late July following Jones's original Facebook announcement. But interest then disappeared - there was virtually nothing about it in the first half of August, remember.

So there was, it seems, nothing inevitable about this story going global. It had a chance to become a big deal in late July - and it didn't. It had another shot in mid-August, and it got a bit of press that time, but then it all petered out.

Only this week has the story become massive. US commander in Afghanistan General Petraeus spoke out on September 6th; ironically, just before the story finally exploded, since as you can see on the Google Trends above, searches were basically zero up until September 7th when they went through the roof.

So the "Koran Burning" story had three chances to become front-page global news and it only succeeded on the third try. Why? The easy answer is that it's an immediate issue now, because the burning is planned for 11th September - tomorrow. But I wonder if that's one of those post hoc explanations that makes whatever random stuff that happened seem inevitable in retrospect.

The whole story is newsworthy only because it's news, remember. The more attention it gets, the more it attracts. Presumably, therefore, there's a certain critical mass, the famous Tipping Point, after which it's unstoppable. This happened around September 6th, and not in late July or mid August.

But there's a random factor: every given news outlet who might run the story, might decide not to; maybe it doesn't have space because something more important happened, or because the Religion correspondent was off sick that day, etc. Whether a story reaches the critical mass is down to luck, in other words.

The decision of a single journalist on the 5th or the 6th might well have been what finally tipped it.

Autistic Toddlers Like Screensavers

Young children with autism prefer looking at geometric patterns over looking at other people. At least, some of them do. That's according to a new study - Preference for Geometric Patterns Early in Life As a Risk Factor for Autism.

Pierce et al took 110 toddlers (age 14 to 42 months). Some of them had autism, some had "developmental delay" but not autism, and some were normally developing.

The kids were shown a one-minute video clip. One half of the screen showed some kids doing yoga, while the other was a set of ever-changing complex patterns. A bit like a screensaver or a kaleidoscope. Eye-tracking apparatus was used to determine which side of the screen each child was looking at.

What happened? Both the healthy control children, and the developmentally delayed children, showed a strong preference for the "social" stimuli - the yoga kids. However, the toddlers with an autism spectrum disorder showed a much wider range of preferences. 40% of them preferred the geometric patterns. Age wasn't a factor.

This makes intuitive sense because one of the classic features of autism is a fascination with moving shapes such as wheels, fans, and so on. The authors conclude that

A preference for geometric patterns early in life may be a novel and easily detectable early signature of infants and toddlers at risk for autism.

But only a minority of the autism group showed this preference, remember. As you can see from the plot above, they spanned the whole range - and over half behaved entirely normally.

There was no difference between the "social" and "geometrical" halves of the autism group on measures of autism symptoms or IQ, so it wasn't just that only "more severe" autism was associated with an abnormal preference.

They re-tested many of the kids a couple of weeks later, and found a strong correlation between their preference on both occasions, suggesting that it is a real fondness for one over the other - rather than just random eye-wandering.

So this is an interesting result, but it's not clear that it would be of much use for diagnosis.

Pierce K, Conant D, Hazin R, Stoner R, & Desmond J (2010). Preference for Geometric Patterns Early in Life As a Risk Factor for Autism. Archives of general psychiatry PMID: 20819977

The Horror, The Horror

You're watching a horror movie.

The characters are going about their lives, blissfully unaware that something horrifying is about to happen. You the viewer know that things are going to end badly, though, because you know it's a horror movie.

Someone opens a closet - a bloody corpse could fall out! Or they're drinking a glass of water - which could be infected with a virus! Or they're talking to some guy - who's probably a serial killer! And so on.

The effect of this - and a good director can get a lot of mileage from it - is that scenes which would otherwise be entirely mundane, are experienced as scary, purely because you know that something scary is going to happen, so you see potential horror in every innocent little thing. An expectation as to what's going to happen, leads to you interpreting events in a certain way, and this creates certain emotions.

In a medical context, that would be called a placebo effect. Or a nocebo effect when expectations make people feel worse rather than better.

The horror movie analogy is useful, because it shows that placebo effects don't just happen to other people. We all like to think that if we were given a placebo treatment, we wouldn't be fooled. Unlike all those silly, suggestible, placebo responders, we'd stay as sick as ever until we got a proper cure.

I wouldn't be so sure. We're always interpreting the world around us, and interpreting our own thoughts and feelings, on the basis of our expectations and beliefs about what's going on. We don't suddenly stop doing this when it comes to health.

Suppose you have the flu. You feel terrible, and you're out of aspirin. You don't think you'll be able to make that meeting this afternoon, so you phone in sick.

Now, clearly, flu is a real disease, and it really does make you feel ill. But how do you know that you wouldn't be able to handle the meeting? Unless you have an extensive history of getting the flu in all its various forms, this is an interpretation, a best guess as to what you'll feel in the future, and it might be too pessimistic.

Maybe, if you tried, you'd get on OK. Maybe if you had some aspirin that would reassure you enough to give it a go. And just maybe it would still have worked even if those "aspirins" were just sugar pills...

Link: See my previous posts I Feel X, Therefore Y and How Blind is Double Blind?

A PCR Primer

The latest episode in the nail-biting scientific drama of "Does The XMRV Virus Cause Chronic Fatigue Syndrome?" has arrived, in the form of a paper in PNAS. A team of virologists led by the renowned Harvey Alter reported finding various XMRV-like viruses, but not XMRV itself, in chronic fatigue patients.

There's been plenty of excellent coverage of this new study, but most of it has come from specialists and has assumed a certain degree of technical knowledge. So here's my attempt to provide a summary for the non-expert, writing as someone who last got his hands dirty in a molecular lab 5 years ago...

The key to the controversy is PCR, a very useful technique invented by a guy on acid (kind of). PCR means Polymerase Chain Reaction. Polymerase is an enzyme which copies DNA. If you ask it nicely, it also copies the copies, then copies the copies of the copies, and so on. Thanks to this chain reaction, you can start with a tiny bit of DNA and end up with loads.

Using PCR, you can detect certain DNA sequences, for example, the DNA sequence of XMRV. (XMRV itself has RNA, rather than DNA, but as a retrovirus, it's able to insert itself into the DNA of infected cells.)

Here's how. DNA is a chain, or strand, of simple molecules called nucleotide bases. There are four: A, C, T, and G. Most of the time, DNA molecules are double-stranded, containing two chains of bases paired up (bound) together. Whenever one strand has A, the other has C, and vice versa. T and G pair up in the same way. They can only pair up in that particular way. T can't pair with C, or G, or with another T.
PCR takes double-stranded DNA and makes more of it. It does this by taking each strand and adding a second strand which is the "opposite" (complementary) sequence of the original, with T and A swapped, and C and G swapped.

That's nothing more than a replica of the original double-stranded DNA.

However, there's a catch. Polymerase can't start a strand of DNA out of nothing, it can only make an existing strand longer. So it needs a primer which can bind to the original DNA and provide "something to work with".

No primer, no duplication. The primer has to be specific: it has to be able to pair up with the DNA. This fact allows us to use PCR to detect specific DNA sequences. Suppose you want to know whether a sample of DNA contains a certain gene, and you know that this gene starts with AAAAA, and ends with CCCCC.

You would make some corresponding primers: a forward primer AAAAA and a reverse primer GGGGG. If the gene is present, these primers will bind to the corresponding target sequences bookending the gene of interest. The PCR will work, and you'll end with loads of copies of that gene. Hooray. If not, nothing much happens. Note that the forward primer is the "opposite" of what you might expect, because it has to bind to the complementary strand. The two primers bookend the region to be amplified - see this pic for an explanation of why.

Once you've run the PCR it's relatively easy to tell whether it amplified the gene or not. But remember that PCR doesn't detect genes, it detects primer targets. The DNA in between the target regions could be anything, as long as the primers fit. In fact, you can tell the length of the amplified DNA, which does provide some information. You can also resequence the amplified DNA to see exactly what it is, but that's expensive.

On the other hand, the match has to be exact. If you're testing for a gene starting with AAAAA, and that gene is present except that it starts with AAAAC instead, you won't find it: a single base difference in the primer sequence throws the whole thing off.

So if someone "used PCR to detect dog DNA", what they mean is that they used primers which they think are specific to dog DNA. This relies on two things being true: that the primers do in fact match the DNA of all dogs (not just some breeds of dog) and that they only match dog DNA (not cats, or mice.)

There are also technical considerations. PCR is vulnerable to contamination by unwanted DNA, because it's so sensitive: even a tiny bit of contamination will cause a false positive. Rogue DNA could come from anywhere: from the researcher running the experiment, from other samples in the lab... So, every PCR experiment needs a negative control, a sample known not to contain the gene of interest. A drop of water is the simplest example. If you "detect" the gene in the negative control, you have to try again (after cleaning all your equipment and washing your hands.)

PCR also doesn't always work. It's like cooking: you have to have the right mix of ingredients, the right temperature, the right timing. If not, you'll end up with a mess. This is why every PCR experiment needs a positive control, i.e. a sample in which you know the gene of interest is present. If you fail to detect the gene in the positive control, you have to check the recipe and try again.

How does this relate to the XMRV story? That's another post...

Normal? You're Weird - Psychiatrists

Almost everyone is pretty screwed up. That's not my opinion, that's official - according to a new paper in the latest British Journal of Psychiatry.

Make sure you're sitting down for this. No less than 48% of the population have "personality difficulties", and on top of that 21% have a full blown "personality disorder", and another 7% have it even worse with "complex" or "severe" personality disorders.

That's quite a lot of people. Indeed it only leaves an elite 22.5% with no personality disturbances whatsoever. You're as likely to have a "simple PD" as you are to have a normal personality, and fully half the population fall into the "difficulties" category.

I have difficulties with this.

Where do these results come from? The Adult Psychiatric Morbidity Survey, which is a government study of the British population. They phoned up a random sample of several thousand people, and gave them the SCID interview, in other words they asked them questions. 116 questions in fact.

48% of people answered "yes" to enough questions such that, according to their criteria, they had "personality difficulties". They defined "personality difficulties", which is not a term in common use, as being "one criterion less than the threshold for personality disorder (PD)" according to DSM-IV criteria.

So what? Well, as far as I'm concerned, that means simply that "personality difficulties" is a crap category, which labels normality as pathological. I can tell that most of people with "difficulties" are in fact normal because they are the literally the norm. It's not rocket science.

So we can conclude that "personality difficulties" should either be scrapped or renamed "normal". In which case the weird minority of people without any such features should be relabelled. Maybe they are best known as "saints", or "Übermenschen", or perhaps "people who lie on questionnaires".

This, however, is not what the authors say. They defend their category of Personality Difficulties on the grounds that this group are slightly more likely to have a history of "issues" than the elite 22.5 percent, e.g. homelessness (3.0% vs. 1.6%), 'financial crisis' (10.1% vs. 6.8%), or having had treatment for mental illness (11% vs 6%).

They say:

The finding that 72% of the population has at least some degree of personality disturbance is counterintuitive, but the evidence that those with ‘personality difficulty’ covering two out of five of the population [it's actually closer to half], differs significantly from those with no personality disturbance in the prevalence of a history of running away from home, police contacts, homelessness... shows that this separation is useful from both clinical and societal viewpoints.

Well, yeah...but no. The vast majority (90+%) of people with Personality Difficulty had no history of these things. It's true that, as a group, they have higher average rates, but all this tells you is that some of them have problems. I suspect they're the ones right at the "top end" of this category, the people who are almost into the next category up.

Here's what I think is going on:

The "difficulties" group and the "none" group are essentially the same in terms of the levels of crap stuff happening to them - because they are the same, normal, everyday people - except that a small % of the "difficulties" group do have some moderate degree of problems, because they are close to being "PD".

This does not mean that the "difficulties" category is good. Quite the reverse, it means it's rubbish, because it spans so many diverse people and lumps them all together. What you should do, if you insist on drawing lines in the sand, would be this:

Now I don't know that that's how things work, but it seems plausible. Bearing in mind that the categories they used are entirely arbitrary, it would be very odd if they did correspond to reality.

To be fair to the authors, this is not the only argument in their paper. Their basic point is that personality disturbance is a spectrum: rather than it being a black-and-white question of "normal" vs."PD", there are degrees, ranging from "simple PD" which is associated with a moderate degree of life crap, up to "complex PD" which has much more and "severe PD" which is worst of all.

They suggest that in the upcoming DSM-V revision of psychiatric diagnosis, it would be useful to formally incorporate the severity spectrum in some way - unlike the current DSM-IV, there everything is either/or. They also argue that with more severe cases of PD, it is not very useful to assign individual PD diagnoses (DSM-IV has no less than 10 different PDs) - severe PD is just severe PD.

That's all fine, as long as it doesn't lead to pathologizing 78% of the population - but this is exactly what it might do. The authors do admit that "the SCID screen for personality disorder, like almost all screening instruments, overdiagnoses personality pathology", but provide little assurance that a "spectrum" approach won't do the same thing.

Yang M, Coid J, & Tyrer P (2010). Personality pathology recorded by severity: national survey. The British Journal of Psychiatry 197, 193-9 PMID: 20807963