[LINK] Antidepressants: Bad Drugs... Or Bad Patients?

In my view, the authors miss out on the real problem with recruiting depressed people through adverts: depressed people don't tend to respond to adverts, because depressed people don't do anything. That's why they call it depression.

Not exactly. Depression includes both misery and inertia, and they're somewhat independent from each other. There are people who get a good bit done, but are miserable-- sometimes suicidal-- for no apparent reason. There are people who can enjoy low-effort hobbies, but have a very hard time doing much of anything else.

Could similar mechanisms hold true for other drugs?

I can see that being an issue for subjective side-effect reporting (e.g. headache, nausea), but not for objectively measurable data, such as blood cholesterol levels, survival rates, etc. There you may have the "true" placebo effect, but probably not the described selection effects.

This is especially true for antidepressants because some are only effective on more severe cases (eg Zoloft); self-selection will yield a body of faux-depressed and mildly depressed people on whom the drug has no result.

Could similar mechanisms hold true for other drugs?

Hmm. This issue could actually be quite broad, and taint a lot of pharmaceutical research data pretty badly. How many people do you suppose there are, who go around signing up for every paid trial they can find? They'd have to defraud the researchers, of course; no one would allow a patient like that into their study knowingly. And what do you suppose that sort of person would do, and how would it be reflected in the data?

Well, if you were defrauding researchers to get them to pay you for participating in their studies, you wouldn't actually take the drugs when they weren't looking. Normal risk from a drug trial turns into a major risk when you're in a bunch of trials at once with possible interactions. But you wouldn't want to make that obvious, because then you might get caught. So you'd start by figuring out whether you were in the placebo group or not, either by trying the pills once and seeing if there was an effect, or breaking them open and seeing if they taste like sugar, or something like that. If in the placebo group, you'd answer all questions consistently with getting no effect. If in the experimental group, you'd try to match the experimenters' expectations, to avoid attention; which would mean making up generic side effects, and claiming some benefit.

Now, how many fraudsters like that would it take, to ruin a study? Since studies count as successful with small effect sizes, a small percentage will do. Now, how many fraudsters are actually in studies, and do they act like I just described?

I propose a "study on study fraud". It's advertised as an antidepressant study, is conducted like one, and pays well. However, all subjects get sugar pills, in a special bottle that looks normal but which contains electronics that log every time it's opened, and which weighs its contents when it's closed. Anyone who empties the bottle all at once is a fraudster. This would yield a measurement of the fraud rate, a list of names to strike from other studies and re-analyze the data, and some survey and interviews which show what fraudsters' responses look like.