28 Comments

The concept of regarding a placeo group as being akin to a control group is fundamentally flawed. A control group - a notion that is very applicable to the physical sciences - has no natural equivalent in psychiatry or psychology, due to unavoidable social interactions.

To elaborate - the placebo group is not just swallowing pills, but is also receiving positive attention, and therefore social support, from those who administer the treatments to them. People are showing that they care. This could make a huge difference, and i would suggest the social solidarity that is evident in the actions of the non-patients in these trials is the real placebo.

So we really have three factors at work in these trials:

1. anti-depressants2. placebos3. social solidarity

Both groups are receiving significant doses of either 1 and 3, or 2 and 3, and therefore the scientific ideal of altering one variable while holding all others constant is not obtained.

Expand full comment

Jeff,would you suspect that it's the color, or the size of the placebo that makes one better than the other?

Do you think that competitive pressures will result in all placebos eventually being, say, big and green?

Expand full comment

No, the benefit is forcing the public to face the fact that we don't have nice tidy solutions for depression and making them tackle the hard decisions like whether to use mood boosting agents (long considered not really anti-depressants since they treat more than depressed people).

More fundamentally the failure of antidepressants represents a serious challenge to the disease model of depression. We need to start taking seriously the possibility that depression is the bad end of a distribution. Depression may be something much more like height or amount of body hair, a natural variation (affected by environmental factors) which we think requires treatment in some cases, e.g., excessive body hair or severely inconvenient shortness.

This issue poses a fundamental moral dilema to the human species. If depression can't be easily fenced off as abnormal brain function when do we use psychoactive substances to alleviate suffering? I think the answer is whenever they prevent more suffering than they cause and the myth of antidepressants is one of the things that is holding us back from developing useful agents of this kind.

Finally saying people are individuals and vary is find and dandy but that still doesn't provide you evidence the drugs are a net help. Heck, those individuals differences could just as well work out in the other direction. Now certainly there are some areas where despite good evidence from controlled studies we can reasonably assume that a given treatment is likely to be helpful to certain patients on the basis of a sound theoretical model but that is distinctly lacking for anti-depressants.

Ultimately though the question isn't whether anti-depressants help some people but WHY. Is it a placebo effect or some other mechanism. Personally I suspect the remaining statistically significant effects for the very depressed result from regression to the mean (if you are way on the unhappy side then randomly fuck with your brain function and your more likely than not to end up a bit happier).

But Robin's point still stands. It's not an excuse for poor reasoning.

Expand full comment

Speaking as someone who has suffered form depression that's no excuse to insist that they aren't placebos. Of course if you don't have bad habits of reading up on the medical literature for drugs you take like I do that placebo effect might do you some good so go ahead and take em.

As for what else you could do. Well if the pyschiatric community really faced this problem they might start prescribing drugs with clear anti-depressant effects like Adderall. Sure there is abuse potential but that's not a reason not to give a person a shot at happiness or to pull them out of a suicidal spiral.

Expand full comment

At some point, Robin, you need to start asking what the expected value of your posts are. This one has some chance of saving someone from a needless expense of $20-$200 to try out antidepressants, and some other chance of encouraging someone to go off their meds, after which they kill themselves or someone else. My estimate is that the expected-value ratio here is many orders of magnitude in the wrong direction. I wonder if you're going to attack vaccination of children next.

Controlled studies are not and should not be the sole arbiter of truth.Controlled studies are especially important in healthcare, because biases, illusions, wishful thinking, and placebo effects are so strong. But they also fail most dramatically in healthcare, mainly because individual reactions differ so much. The assumptions underlying standard statistical analyses don't apply to human biology or psychology, which never have the simple distributions that the analysis requires.

When one person can repeat something many times and have an observable effect each time, then we need to take their anecdotes seriously. For instance, I get a headache from drinking two sodas with Aspartame in a row. Any respectable MD will swear up and down that this is impossible, because many controlled studies have shown that Aspartame has no ill effects. Nonetheless, my own personal experience is repeatable and reliable.

In other cases, say orthoscopic surgery, where an individual doesn't get repeated trials, we don't have any way to observe individual variation other than in large studies. In that case, believing the results of large studies is more appropriate than believing the impressions of doctors who have never bothered to actually count how many patients said they felt better afterwards.

Expand full comment

I do believe that a number of antidepressants have been shown to help a lot with suicidality, but not with the overall depression. That's definitely my experience.

Expand full comment

Menand's article points this out too:

Kirsch’s claims appeared to receive a big boost from a meta-analysis published in January in the Journal of the American Medical Association and widely reported. The study concludes that “there is little evidence” that antidepressants are more effective than a placebo for minor to moderate depression. But, as a Cornell psychiatrist, Richard Friedman, noted in a column in the Times, the meta-analysis was based on just six trials, with a total of seven hundred and eighteen subjects; three of those trials tested Paxil, and three tested imipramine, one of the earliest antidepressants, first used in 1956. Since there have been hundreds of antidepressant drug trials and there are around twenty-five antidepressants on the market, this is not a large sample. The authors of the meta-analysis also assert that “for patients with very severe depression, the benefit of medications over placebo is substantial”—which suggests that antidepressants do affect mood through brain chemistry. The mystery remains unsolved.

Read more: http://www.newyorker.com/ar...

Expand full comment

...anti-testimonials to drugs that definitely do nothing, such as homeopathic remedies, are rare. Anti-testimonials to antidepressants are common.

Isn't that because homeopathic remedies do nothing, but antidepressants screw with your brain chemistry? Just because the antidepressants have bad side effect doesn't make Horn 2 particularly more likely (unless you say "well, it screws with my brain, and my depressions is inside my brain, therefore it's more likely to be effecting my depression than if I had no side effects".)

Expand full comment

Is Robin Hanson advocating CBT (=Cognitive behavioral therapy?)

Expand full comment

The problem with this is that a placebo effect isn't necessarily a real cure, and I would contend that it is unlikely that there are any real effects from placebo. Placebos control for all effects other than the active effects of the drug - most notably, I would say, reporting artifacts. If someone is more likely to report that they feel better even if they don't feel better (possibly due to imperfect ability to compare their current state to past states), this will show up as a "placebo effect" in people who are objectively no better (and would be objectively no better if an objective measure existed).

I would be willing to bet that you could find placebo effects in treatments of a rock to increase shininess if you performed such a study.

Expand full comment

That's only if the patients know what side effects to expect.

You probably can't get away with giving them a 'mystery drug', but you might be able to add a few extra "possible side effects" to the list. If you can do that, then just add on whatever side effects your new placebo has. Informing people of negligibly likely risks is standard practice anyway.

Expand full comment

whether or not these drygs work, and i think some of them do (particularly citalopram) help to relieve frustration and anger, i just don't feel the side-effects are worth it

i'll tell you this, i took citalopram for 3 months. for the entire first month i constantly felt like i was coming up on ecstasy. god knows what i would've thought was happening to me if i had no experience of illegal drugs.

that and the fact that i was basically sedated, felt no anger at the usual unreasonable things that make me angry (abusive customers, fascist street gangs marching in my country, etc) and was basically impotent for the entire period has convinced me never to touch psych drugs ever again.

and you'd better believe that if you've basically been high on E for 3 months, the come-down is HARD, and it lasts for weeks

again, luckily, i've got experience with drugs, and i know what to expect. someone who didn't might very well find themselves in very dangerous and unknown psychological territory

that's not to say that properly prescribed anti-depressants might not be helpful for people with severe depression. but over here they hand the things out like sweeties

Expand full comment

Don't most of these trial have retrospective queries, so that participants can tell the trial which arm they thought they were in? This allows for the "placebo-reveal" effect to be determined. Thus you could throw out a trial in which everyone correctly guessed which arm they were in. This is fairly standard practice currently, I don;t know if some of these older trials used this.

Expand full comment

At the end of your Double Blind Randomised Controlled Trial you open the envelopes and half of your control group died. Sad, but typical. Pressing on, you find that half of your treatment group died. That is a huge disappointment: the drug did nothing at all!

That is the usual interpretation, but it is wrong. It may well be that the drug did nothing at all, but the result of the trial is logically weaker. The drug did nothing OR it re-rolled the dice, curing some who would have perished and killing an equal number who would have recovered on their own.

That leaves the medical researcher with a dilemma. Horn One: reject the hypothesis that made the trial seem worth running and go in search of a different drug with a new mechanism. Horn Two: reject the hypothesis that the patient population is homogeneous and go in search of a diagnostic test that will tell you in advance who you will kill and who you will cure.

What extra information might make Horn Two look best? Testimonials are obviously consistent with the drug curing some patients who would otherwise have perished, but testimonials are in practice quite consistent with drugs that do nothing at all. Anti-testimonials are more interesting, because anti-testimonials to drugs that definitely do nothing, such as homeopathic remedies, are rare. Anti-testimonials to antidepressants are common.

What should a depressed person do if they accept that antidepressants are a feeble medicine overall, but they also accept the package of testimonials and anti-testimonials as pointing strongly towards Horn Two? It is reasonable to believe the testimonials and try taking the drugs, but that is the playing-with-fire option. Given the disappointing overall results, the more you believe in the testimonials, the more you should also believe in the anti-testimonials, and you should have a correspondingly greater wariness that you might be harmed rather than helped.

Expand full comment

Agreed. Robin, in his identity as against medicine, has forgotten that placebos, like drugs, are different. All those "eloquent authors" are saying that this placebo is better than those other placebos.

Expand full comment

Eliezer you need a placebo that mimics the side-effects of the active drug. - so your placebo has some active ingredients as well, which complicates the study (perhaps those ingredients have more side effects..)- and you'd need the cocktail to be licenced for the use you intend- and if the side effects you are mimicing are harmful there are ethical problems. eg how do you test chemotherapy against a placebo?

it's kind of complex.

Expand full comment