News Round-Up: September 2009 4th Edition


Research on Antidepressants

The British Journal of Psychiatry features two interesting studies on antidepressants. The first features a patient rating scale for antidepressant side-effects – the Antidepressant Side-Effect Checklist (AEC) which is included in the Appendix for the paper (Uher et al, 2009). The researchers compared this patient rating scale with a clinician rating-scale, the UKU in 811 subjects with depression who were participating in an open-label trial comparing Nortriptylline with Escitalopram. The Nortriptylline was included because of a strong affinity for noradrenergic receptors (it would have been interesting to see whether similar findings would have occurred with Reboxetine). They found that after correcting for the severity of depression, the AEC scores predicted discontinuation of escitalopram (although curiously not the Nortriptylline) and validated the use of the instrument for the purposes of establishing side-effects in antidepressants. In another study, this time qualitative, the researchers explored the emotional side-effects of the SSRI’s. The responses from the participants were grouped into 7 categories and there were many interesting comments from the participants (Price et al, 2009). Both a reduction in ‘positive’ and ‘negative’ emotions were reported and there was some supporting evidence from an analysis of comments on several depression related online forums. The authors suggest further quantitative studies to investigate the findings from this study.

Financial Skills and Risk of Dementia

Predicting which people with Mild Cognitive Impairment go on to develop dementia is an area of current research interest. There are many studies using different methodologies looking into this question. One predictor is that the size of the Hippocampus (size is inversely correlated with dementia risk) which has a robust evidence base. However, a recent study provides evidence that financial skills may be another marker of risk and this has been widely reported in the media (e.g. here, here and here). A research team, just published in ‘Neurology’ found that people with Amnestic Mild Cognitive Impairment who scored poorly on the Financial Capacity Instrument were more likely to develop dementia. The sample group were people with Amnestic MCI and are therefore already a select group who have been assessed as having formal difficulties with memory. They were being scored on a tool which measures financial skills. The size of the study is relatively small (n=163) and of these, 25 people with Amnestic MCI went on to develop dementia.

Blog Round-Up

Just to mention that the dead salmon study is generating a buzz in the blogosphere. The question is why the salmon’s brain activity showed up on the fMRI scan. Three issues affecting the interpretation of fMRI data are described in Developing Intelligence’s article – multiple comparisons, cluster comparisons and independence of multiple tests. For the interested reader, the pdf of the poster is located here. The authors presented the salmon with photographs followed by a ‘rest’ period and then analysed the change in BOLD signals in voxels. They performed one analysis without corrections. This ‘revealed’ a region of activity in the Salmon’s brain. They then carried out two types of comparison that are routinely used in the analysis of fMRI data using software packages such as SPM2. These corrections are necessary because if there is ‘random noise’ then some voxels should be firing above the threshold level just by chance. Thus you need to correct for both the combination of chance and the large number of voxels that are being examined. The two methods that are routinely used and included as a default in one statistical package are family wise error rate correction and false discovery rate correction.The debate that is currently taking place revolves around some posts which have likened these results to those of another study originally called ‘Voodoo Correlations in Social Neuroscience’ (reviewed here) and provoking much scandal before it surreptitiously changed its name to ‘Puzzling High Correlations in fMRI Studies of Emotion, Personality and Social Cognition’ (reviewed here). The point made in the Developing Intelligence article is that the ‘Voodoo Correlations in Social Neuroscience’ paper focused on independent tests whereas this paper focuses on multiple comparisons and thus a number of bloggers have misrepresented the paper. I would argue that both sides are probably right. Thus (and at the risk of sounding slightly pedantic), the Voodoo/Puzzling paper takes a broad swipe at fMRI study methodology from the statistical analysis right the way through to the way in which the analysis is presented within the papers. Thus

1. They do explicitly at one point, mention within-subject comparisons (i.e. changes occuring between tasks in the same subject)

2. In the second part of the discussion, the authors specifically ask the rhetorical question

‘Is the problem being discussed here anything different than the well known problem of multiple comparisons raising the probability of false alarms?’

In their response, the authors state that ‘The problem we describe arises when authors then report secondary statistics on the data in the voxels that were selected originally‘ before then adding ‘In the case discussed in this article, correlations are both the selection criterion and the secondary statistic‘.

3. However, most telling is Vul et al’s response to the responses to their original paper (available here) where they state explicitly that

6. We pointed out that although most studies used appropriate multiple comparison corrections (and thus identified voxels that do indeed have nonzero correlations), these methods are not always correctly applied, even in articles that have passed peer review’

One salmon tries to leap over a large waterfall while the other jumps into a hungry bear's mouth
One salmon tries to leap over a large waterfall while the other jumps into a hungry bear's mouth

However, taking the study a little further in a slightly tongue-in-cheek fashion, one interesting question to ask is

‘Why was any activity being picked up at all?’

The fMRI scan typically picks up oxygenated-Haemoglobin and it is the change in oxygentation of the Haemoglobin with delivery to the brain tissue that produces the signal. So if the Salmon was dead, we could safely assume that there was neither blood flow nor any delivery of oxygen to the brain tissue. So where is the change coming from? Well, I thought there was something fishy going on here and I was unsurprised to find a paper on the analysis of Salmon blood (freely available here). After looking very briefly at this paper, it appears the researchers in the fMRI study have omitted some crucial details for the interpretation of the data. Thus we do not know if their selected Atlantic Salmon was caught in fresh water or bracken water, as the corresponding temperatures would affect the oxygen dissociation curves of the salmon’s haemoglobin. We also need further information on the temperature of the room at the time of the scanning as well as the period of time for which the salmon had been deceased.

There is one possible explanation for the BOLD signals. They could represent a dissociation of the oxygen from the Haemoglobin with time. When the dissociation occurs, the magnetic properties of the Haemoglobin are altered and this will be picked up by the MRI scanner (oxygenated Haemoglobin is diamagnetic and deoxygenated Haemoglobin is paramagnetic). On moving from the presentation of the photographs through to the rest period there may be a continuing dissociation of oxygen from the Haemoglobin and it is this which is being picked up by the scanner. More interesting (and a very remote possibility) is that the authors of the poster have picked up deoxgenation in the Salmon’s Substantia Nigra, a region which is associated with Parkinson’s Disease in humans. The reason I mention this region is two-fold. Firstly, on the admittedly low resolution image of the Salmon’s brain on the poster, in the sagittal section, the activity is located close to the spinal cord and may therefore correspond to the midbrain (although I know next to nothing about Salmon neuroanatomy. I did however find a reference to the Substantia Nigra in Salmon). Secondly, in humans at least, the Substantia Nigra contains Haemoglobin which may offer it some protection during periods of anoxia and which also gives it the has a characteristic dark stained appearance (hence the name Substantia Nigra*). Salmon do a lot of moving about (as in the above picture) and the Substantia Nigra is involved in the regulation of movement. However against this hypothesis, the corrections in the statistical analysis eliminated the activation patterns, and there were at least two areas of activation in the uncorrected analysis.

Human Substantia Nigra - From Gray's Anatomy 20th Edition
Human Substantia Nigra - From Gray's Anatomy 20th Edition

So in summary, there’s a remote possibility that the imaging study was picking up the dissociation of oxygen from Haemoglobin in the Salmon’s Substantia Nigra post-mortem or as they infer, the activity was due to their incomplete analysis (particularly as the differences disappeared after the necessary corrections were made).

* The characteristic dark appearance of the substantia nigra is from the melanin in dopaminergic neurons.


There may be another explanation for the findings. Thus if there is a mix of red blood cells (RBC) of varying levels of oxgenation, then the scanner may causes a realignment of the RBC’s according to their magnetic properties. If the magnetic field is applied as a pulse, then there will be further realignments when the magnetic field is changing. This may produce eddy currents, in which the Haemoglobin generates a reactionary magnetic field which will then ‘resist’ the next pulse of the scanner. If this were the case, it means that Haemoglobin molecules would undergo a more marked change in alignment during the initial part of the scan and this might explain the difference.

On a slightly different note, the brightest light in the known universe – the Diamond Light Synchrotron (see Podcast review here) is being used to examine iron deposits in Alzheimer’s Disease and together with the finding of Haemoglobin in the Substantia Nigra A9 cells, this is a very current area of research.

Addendum (4.10.9)

See the comment below from the author of the poster – Craig Bennett (also author of the blog at on the interpretation of the results.


Price J et al. Emotional side-effects of selective serotonin reuptake inhibitors: qualitative study. The British Journal of Psychiatry. 2009. Vol 195. No 3. 211-217.

Uher R et al. Adverse reactions to antidepressants. The British Journal of Psychiatry. 2009. Vol 195. No 3. 202-210.


You can follow ‘The Amazing World of Psychiatry’ Twitter by clicking on this link


You can listen to this post on Odiogo by clicking on this link (there may be a small delay between publishing of the blog article and the availability of the podcast).


If you have any comments, you can leave them below or alternatively e-mail


The comments made here represent the opinions of the author and do not represent the profession or any body/organisation. The comments made here are not meant as a source of medical advice and those seeking medical advice are advised to consult with their own doctor. The author is not responsible for the contents of any external sites that are linked to in this blog


  1. Justin – Great discussion of the dead salmon results. I just wanted to respond with some additional background on the study and some theoretical comments on the source of the false positives.

    * Regarding the full background of the fish, we will never really know what kind of water it was kept in while alive or exactly how long it had been deceased. After talking to the supermarket last week our best guess is that it had been dead at least 48 hours before it reached the seafood counter.

    * There was a region-of-interest in the dorsal spinal column, but I am not certain if it represents the substantia nigra in the salmon. My ichthyological anatomy is a bit rusty.

    * The possibility of oxygen dissociating from hemoglobin is an intriguing possibility. I cannot discount that this process might be taking place in the salmon during our scanning. However, the design of our fMRI experiment and the image processing steps we take during analysis both provide some evidence against it causing our false positives.

    The experiment was set up as an interleaved block design, meaning that there were several alternating periods of stimulus presentation (photos) and rest (no photos). If the dissociation process were contributing to the results then it would have to alternate in synchrony with the changes in stimuli. There are no changes in the fMRI acquisition associated with the different conditions, so it would have to be due to the presence/absence of the photos.

    Second, during our analysis we apply a temporal high-pass filter to the timeseries to remove any slow drifts in signal. These low-frequency drifts are usually due to small changes in the scanner, such as the gradient magnets warming up. A side benefit is that this filtering would also remove any slow changes in signal due to oxygen dissociation.

    Best ~ Craig [prefrontal]


  2. Hi Craig,

    Thanks for the comments and the great poster. It sounds as if there are two issues here.
    1. Why is there a signal in the first place? If there is no blood flow or transfer of oxygen between the tissues then why is there a signal? Why wouldn’t there just be ‘no signal’ at all. If there was ‘no signal’ then there would be no noise or opportunity for false positives. However, there is a signal and this has to come from somewhere. Maybe this is coming from the scanner itself – but as you’ve pointed out – you’ve used the temporal high-pass filters. The salmon is decomposing. So if you’ve excluded the slow drift, there remains the possibility that there is a valid process occurring which changes the magnetic properties of the Salmon – there are various chemical decomposition processes which might account for this. Admittedly if there was dissociation of oxygen from the Haemoglobin, then the signal would be expected in all perfused areas which wasn’t the case. However activity in other areas might be shown if the experiment was repeated (if it’s due to the statistics).

    2. There second part is why was there a difference between the ‘task’ and rest periods and here there is only one explanation. Of course as you suggest the dissociation of oxygen would occur in both rest and task periods so it wouldn’t explain any differences between those periods. Here the statistics rule – and the analysis must be producing the artefact. But it still needs the input to produce the artefact and that’s the ‘mysterious’ signal.

    Maybe there’s a story behind the signal! Any chance of a follow-up poster? 🙂




  3. Justin,

    Thanks for your reply. I apologize about my own delayed response – I had to finish writing a review paper and was off the interwebs for a while. In response to your questions:

    1. When we are conducting fMRI there is always ‘signal’ in the voxels that we measure. The T2* sequences used in fMRI are sensitive to both the inherent T2 signal of the tissues and the additional relaxation time resulting from disturbances in the local magnetic field. Think of it as a kind of anatomical picture of the brain whose values change depending on small changes in the local magnetic field, like blood flow differences. What we are interested in during fMRI is really just the changes in the signal over time, so we ignore most of the baseline anatomical information in the images. It is likely that the signal that we were detecting in the voxels of the salmon was primarily driven by the T2 signal, with perhaps some other slow field changes due to the processes you describe.

    2. The results that we saw in the salmon brain were likely driven by random noise in the acquisition of the fMRI images. There is some amount of error in the measurements that we take, and in this case they just happened to occur within the salmon’s brain. If we did the experiment over again it is likely we would still see active voxels, but they would occur somewhere else within the salmon’s body. That is why proper correction is necessary – you never know when the random noise will line up with your prediction.

    A follow-up poster is tempting. It has been suggested that we scan a whole group (16) of salmon! As it stands right now we are just trying to get the first poster published as a paper. Then we can discuss next steps. 🙂

    Thanks again!

    ~Craig [Prefrontal]


  4. Hi Craig,

    Thanks very much for taking the time out to explain your thoughts on this. Best of luck with the paper! 🙂



  5. you are really a good webmaster. The website loading speed is amazing. It seems that you’re doing any unique trick. Moreover, The contents are masterwork. you have done a wonderful job on this topic!


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s