Why depression is not a useful or reasonable phenotype for research in clinical psychology, psychiatry, or medicine

[Woolie]

A few people were interested in delving into the thorny issue of depression. Here's a nice place to start. Eiko Fried summarises why depression is not a useful concept for research - in a series of 18 tweets.

Click on the first tweet here and the other 17 will appear underneath:



I will post some comments and translations of some of his points shortly.

 

[Alvin]

But according to some estimates 25% of us have the undiscovered genetic defect that causes depression hence we must take the drugs in vogue which change over time to subdue this defect. How can so many people have the same undetectable defect that can't be studied?
I'll refrain from saying more since i don't want members unhappy with what i say beating down my door again.

 

[Scarecrow]

This link was in one of the subsequent tweets:

http://thepsychologist.bps.org.uk/volume-29/january-2016/depression-more-sum-its-symptoms/

 

[Woolie]

A few nuggets.

Fried points out that the DSM criteria for diagnosing major depressive disorder are largely based on historical precedent and clinical opinion -in other words people 40 years ago thought they saw a certain constellation of symptoms as sort of going together (either from experience or based on their preconceptions or a bit of both), and we're pretty much just gone along with this ever since.

He points out that psychometirc scales of depression just add up the total number of symptoms. But since the symptoms do not all co-occur together, you could get the same score from two very different profiles.

There are some interesting articles mentioned too.

This study interviewed more than 1500 patients diagnosed with major depressive disorder. They looked at the different symptoms used to classify depression, and which ones were most predictive of a depression diagnosis. They also looked at how specific the symptoms were (how good they were at distinguishing between depressed and non-depressed patients). They found that sleep disturbance generally and loss of energy (i.e. fatigue) were the least specific by this definition.

Full text here.

 

[Woolie]

But according to some estimates 25% of us have the undiscovered genetic defect that causes depression hence we must take the drugs in vogue which change over time to subdue this defect. How can so many people have the same undetectable defect that can't be studied?

Not quite sure what you mean, so can't really be unhappy about it!

 

[Adrian]

He points out that psychometirc scales of depression just add up the total number of symptoms. But since the symptoms do not all co-occur together, you could get the same score from two very different profiles.

Has anyone done a cluster analysis on the different answers?

 

[Woolie]

Has anyone done a cluster analysis on the different answers?

Yes, this tweet in the series looks at intercorrelations amongst self-reported symptoms.

Edit: Here is the figure in larger form, followed by a key to the abbreviations:




If you ignore the clusters that are sort of obvious (e.g., weight change and appetite change, different kinds of insomnia, anxiety and panic attacks), the tightest cluster of interrelated symptoms involve sadness, loss of pleasure, mood reactivity, and loss of interest in things.

Then the next layer seems to be energy and concentration

The other symptoms are not at all strongly related to the central core group of loss of pleasure/sadness.

One thing to bear in mind is that all these analyses look at people who've already received a diagnosis of depression. So they tell us only about those people who "made it" into the category. This will of course massively enhance correlations between symptoms, because the cohort has already been shaped by tossing out those who don't achieve a minimum number of symptoms.

The matrices are really telling us more about the way we diagnose than about the "reality" of symptoms out there.

 

[Adrian]

Its not that easy to read, because the symptoms are abbreviated, and I can't seem to locate the paper it came from. Gas is gastric symptoms, Ene is energy (fatigue), Wei is weight. I think Qua might be quality of life?

That's trying to look at the relationship between variables. I was wondering if anyone had taken the question answers as a vector and looked for clusters between different patients. Depending on the distance metric used that can show interesting things. But also worth removing noise like variables as part of clustering process as otherwise distance metrics in the clustering can be mislead.

 

[Hoopoe]

When the goal is successful treatment, what matters is the biology. Since different biology can produce the same or similar symptoms, trying to deduce the biological phenotype from symptoms will often not work.

Of course, some people will challenge the claim that biology is what matters, instead making the counter claim that some illness, for example CFS, can be treated effectively with psychobehavioural approaches (CBT). In the case of CFS we know that this is false. Besides, this logic doesn't make a lot of sense if applied to real world problems or other diseases. Imagine if the general solution for chest pain was a pacemaker. Most people with chest pain don't need a pacemaker, and many don't even have a problem with their heart (instead having perhaps acid reflux which is also called heart burn for a reason). But imagine if the pacemakers-for-everyone people never gave up and kept trying to find a way to make pacemakers work for everyone with chest pain, trying different models and surgical techniques and generally just insisting that the problem is a really tough one to solve.

In the field of CFS, the CBT/GET-for-everyone people are still trying to make it work (or did until recently).

 

[Trish]

I've experienced several GP's 'diagnosing' me with depression and prescribing antidepressants or CBT on one simple sign - if I cried in their presence, I was depressed. End of diagnosis, end of consultation.

It has happened at least 3 times. I was not depressed, I was completely physically drained by the exhaustion of daily life with ME, and the effort of getting myself to the GP and trying to convey this to the GP was too much. Needless to say the treatments were worse than useless.

 

[Arnie Pye]

I've experienced several GP's 'diagnosing' me with depression and prescribing antidepressants or CBT on one simple sign - if I cried in their presence, I was depressed. End of diagnosis, end of consultation.

Been there, done that, had the same reaction several times... And it annoys the crap out of me.

Society in general accepts that women and men are not identical beings who respond to situations in exactly the same way, but doctors seem to have different attitudes.

For me (I'm female, despite my user name) crying is absolutely normal behaviour when pain (for example) is unbearable and I am dismissed and left untreated for year after year after year. So, doctors see behaviour (crying) that I consider normal for women under the circumstances, they decide it isn't normal (it isn't behaviour which is common amongst men), and decide that the only cure for this abnormal behaviour is to prescribe anti-depressants.

I almost expect to hear doctors cry "Why can't a woman be more like a man?". But of course they don't. I don't think there was any malice intended when Henry Higgins said it.

Edit : Typo

 

[Scarecrow]

Thanks for starting the thread, Woolie.

Here's a nice place to start. Eiko Fried summarises why depression is not a useful concept for research - in a series of 18 tweets.

It's worse than I thought.

8/ MDD is highly heterogeneous. 227 unique Sx profiles (SP) exist for MDD diagnosis using DSM-5 criteria. When disaggregating the 9 criteria (e.g. insomnia vs hypersomnia), 16400 SP emerge. Patients differ dramatically in SP. (https://www.ncbi.nlm.nih.gov/pubmed/25451401 ; https://www.ncbi.nlm.nih.gov/pubmed/24886017 ).

10/ MDD sum scores add up not only very diverse Sx, but literal opposites (e.g. insomnia vs hypersomnia; weight gain vs weight loss; appetite again vs loss; psychomotor agitation vs retardation). From a measurement perspective, this is pretty stunning.pic.twitter.com/y9LtjMBvTN

So depression may be an even bigger wastebasket diagnosis than ME/CFS.

In another tweet, Fried claims that the reliability of a MDD diagnosis is "barely above chance level".

It's enough to make you wonder how any research into MDD can yield positive results.

This should make us pause whenever we are tempted to explain away poor ME/CFS research on the assumption that people with depression may have been included. For example in the PACE trial, I've often seen it claimed that improvers in the GET group probably didn't have ME/CFS, either the implication or explicit reason being they must have had depression because the Oxford criteria are such a pile of crap. It may well even be true that people with depression who did not have ME/CFS were included in the trial. But can we actually trust research claims of exercise improving depressive symptoms?

Bad research is just bad research. Why would unblinded treatments and subjective outcomes produce reliable research in people with depression if it doesn't in people with ME/CFS?

 

[Scarecrow]

I've experienced several GP's 'diagnosing' me with depression and prescribing antidepressants or CBT on one simple sign - if I cried in their presence, I was depressed. End of diagnosis, end of consultation.

Been there, done that, had the same reaction several times... And it annoys the crap out of me.

Who among us hasn't had a similar experience? Can't be too many.

For me (I'm female, despite my user name) crying is absolutely normal behaviour when pain (for example) is unbearable and I am dismissed and left untreated for year after year after year. So, doctors see behaviour (crying) that I consider normal for women under the circumstances, they decide it isn't normal (it isn't behaviour which is common amongst men), and decide that the only cure for this normal behaviour is to prescribe anti-depressants.

There was a time I would have agreed with you. Now I'm less certain.

A few years ago, I learned that one of my male colleagues had been diagnosed with depression when he was at university. Several months later he got a revised and this time correct diagnosis - Type 1 diabetes. That's progress in sexual equality, I thought to myself. Instead of treating young women with respect and giving more credence to what they report, GPs have instead opted to lower their opinion of young men.

 

[lansbergen]

If it nothing else it must be hidden depression.

 

[Woolie]

This should make us pause whenever we are tempted to explain away poor ME/CFS research on the assumption that people with depression may have been included. For example in the PACE trial, I've often seen it claimed that improvers in the GET group probably didn't have ME/CFS, either the implication or explicit reason being they must have had depression because the Oxford criteria are such a pile of crap. It may well even be true that people with depression who did not have ME/CFS were included in the trial. But can we actually trust research claims of exercise improving depressive symptoms?

Aye to that, @Scarecrow. Saying it probably worked for the "depressed" people is endorsing the potential usefulness of these inteventions, but only for "the right sorts of people" - ie., not us!

Bad research is just bad research. Why would unblinded treatments and subjective outcomes produce reliable research in people with depression if it doesn't in people with ME/CFS?

Totally! The results are exactly what you'd expect from interventions that do nothing but introduce reporting biases - there's not need to bang on about how "depressed people" were the problem.

 

[Woolie]

That's trying to look at the relationship between variables. I was wondering if anyone had taken the question answers as a vector and looked for clusters between different patients. Depending on the distance metric used that can show interesting things. But also worth removing noise like variables as part of clustering process as otherwise distance metrics in the clustering can be mislead.

Not ideal, but I think you can get a bit of a flavour of that from the symptom correlation matrix - that green figure above - because it tells you which symptoms are consistently endorsed by the same people, and which ones vary across people (I understand that to be a vector diagram too, but I'm not so good on those mathematical things).

In the post above, I've added some of my own thoughts from looking at the matrix.