*talks

Adding video to channels (Talks) - requested by eric3579.

Okay... I haven't finished the video, but I'll give you a play by play anyway. It's not my field (I study sensory systems) so if there are any clinicians in the audience, please forgive my ignorance.

First plot: The effect she's talking about is an extrapolation from a linear model with presumably assumed uniform variance along the independent axis. If you look at the actual underlying data, presumably notated with open and filled bubbles (no mention of what size means, but probably number of samples) you see that there is a lot of "effect" extrapolated from the model despite very little obvious trend in that part of the data. In fact, there are only two or three open bubbles at all on that part of the chart presumably because it was viewed as unethical to treat severely depressed patients with placebos. Further, there is a huge (relatively speaking) variability in the efficacy among severely depressed patients treated with the drugs. This is a symptom of one of the root problems which is that clinical modeling is typically very weak. The models are often simplified not because it is appropriate or useful, but rather because it is the way that particular researcher knows how to model data.

The a priori linkage between score on some survey and a response to some chemical is, IMO, tenuous at best. Given an argument from correlation that seems to lean heavily on a magic data point or two does little to change it. To give you a sense of how tenuous this data is, consider instead fitting the lines through only the data where the effect is considered "large." As an experiment, pull up the chart, and do a "chi by eye" fit of a line through the white dots, and a line through the dark dots, but only those dots to the right of the "severe depression" threshold line. Notice that the white line would then have a much steeper slope (stronger correlation) and the dark line would have almost no correlation (that data looks pretty isotropic to me).

Alright... that's enough for one comment. Next plot please...

Second plot, again the only "clinically significant difference" was in a part of the plot where there are very few data points. Arguably insufficient data to believe in the model from which you are drawing the conclusion. While there does seem to be a larger difference in this plot among the patients hovering around 24 or 25, the scale on the y axis has changed substantially (been compressed). It isn't clear that difference is "anything to write home about" so to speak.

Man... I wish she would put error bars on these things. I have no idea what variability in a GAS looks like. Worse, it's very hard for me to assess what clinician bias can do to these data. For example, what the person discussing "side effects" of a placebo might insinuate could have a relatively substantial impact on these data. Interpreting data of these sort is always a challenge.

I'm honestly more impressed with her anecdata. "Most of us know." "Anyone working with patients can see." That sort of stuff. Clearly she's worked with people, and has been a successful "healer." I'm just not yet convinced chemicals have as much to do with it as relationships with therapists and patient compliance. The data just don't seem to speak to that. In fact, little things like the lag in the efficacy she mentioned provide evidence against our understanding of how these chemicals work.

I agree with her though that the problem is diagnostics. I'm just more hopeful about other treatment paradigms in the presence of greater specificity of diagnosis. I might be wrong, I just think that too much is made out of our biochemical breakdowns of the nervous system. It's like caring what doping was put in the transistors in your computer. Yes, you'll find that particular doping agent all over the damn thing, but if the computer is broken, you might want to talk to a programmer not a chemist. It's the software, the particular pattern of ones and zeros (or in the brain analog, the wiring and action potentials) that are usually faulty. Not widespread deficiencies in the functioning of doping compounds.

"The treatments work, and if you have someone very skilled and experienced performing them they work even better."

My response:

"The treatments work if you have someone with clear incentives to demonstrate their efficacy."

That seems fair given she would discuss the bias of people skeptical of the efficacy of the treatments.

I think this is due to doctors prescribing anti depressants for anything. I'm not medical in any way.

I think depression is very different from "sadness for something". I've suffered badly with depression all my teenage and adult life, and seen people prescribed anti depressants because they're sad that they broke up with their boyfriend, or their dog died or something. I think that's a necessary sadness, something you need to go through and you will get through and be better for it. That's what being human is about.

If you don't have a reason to be sad, if no thing or experience could cheer you up, if there was no goal and no desire because nothing held meaning to you; that's depression.

There may be other types, but that's the one i know and i think it can benefit from medical help. At least, i've had medical help from a good doctor (as well as medical help from a bad doctor).

My beef is every person I know who's on anti-depressants has the same story that they went through 3 or 4 before they found the right one for them. Every person. Why didn't the second one work? Or the first?

Seems that she makes the opposite point she's trying to argue. The studies show the antidepressants are no better than placebo in mild to moderate depression. Therefore, they don't work, OTHER than functioning as a placebo. Yes, we all know the placebo effect is real, about 20% of people improve with a sugar pill. So, I guess she's arguing that the antidepressant is just a substitute for the sugar pill, and therefore effective in that regard. Why not save expense and side-effects and just give patients the sugar pill instead, when they have mild to moderate depression? Or better yet, prescribe exercise, bright light therapy, nutrition, abstinence from CNS depressants, and healthy sleep habits instead.
The evidence for a physiologic/anatomical cause of the symptoms of depression is overwhelming. The vegetative symptoms of severe depression; generalized anxiety, appetite disturbance and, the hallmark, sleep disturbance, all point to an abnormal "hypervigilant" state probably mediated through the hypothalamic, pituitary, adrenal axis. The genetic predisposition for depression and the evidence that permanent changes in the brain may progress with frequent bouts of severe depression also point to a common physiologic pathway. Having experienced the acute onset of 2 bouts of severe depression, complete with all the vegetative symptoms, I can tell you from personal experience that you KNOW something has changed in the way your brain and body are functioning. It's if an alien has invaded your body, taken control and tucked your old "self" away in a closet somewhere. I couldn't say that any of the antidepressants that I tried had a profound effect on recovery, it was 6 months of hell, and then about 2 years of discomfort before remission in both cases. However, now I think I'm on a combination drug therapy that I feel is more effective in keeping me in remission. Part of the issue may have been that I'm more bipolar with mainly depressive mood, rather than a true unipolar depressive type.