One researcher on the placebo effect said that the popularity of SSRI anti-depressants represents "the triumph of marketing over science".

American Journal of Psychiatry

This is off topic:
AMERICA
Please! STOP HURTING YOUR CHILDREN!!

I absolutely agree: --"when they start tossing around assertions like "major depressive disorder in children and adolescents is common,' as they do in their paper (pdf below), I tend to get a wee bit testy."-- Current psychiatric research papers are full, FULL, full I say, of unsupported, unquestioned assertions like this.

I'm a psychiatrist, and it is embarrassing that the editorial boards of supposedly scholarly journals let stuff like this pass.

As I continue to say, I love the way you report about the sad state of things. It gives me strength to keep up my protests and rants on my own turf.

Yeah man. It appears to be a pretty substandard and shitty paper that has found a correlation that is probably meaningless. Multiple sites boosting placebo response? Smells like bullshit.

My only gripe with this article, and it's pretty minor, is about the 2008 Arif Khan study in the Journal of Psychiatric Research (42:791-796). It did not show that the placebo response rate was 79% and the antidepressant response rate was 93%. The response rates were 35% for placebo and 51% for antidepressants (p. 795). Of those who responded to placebo and continued on placebo, 79% did not relapse from 16 to 52 weeks, compared to 93% who did not relapse during that time period while continuing on antidepressants. Also, the effect size is not 0.14, because you do not calculate effect size by subtracting two values, but have to divide the difference by the standard deviation. But, like I said, that’s just a minor point. :)

Great job, Philip.

Sorry, just one more thing about your effect size calculations. Effect size is not a percentage value, and so your saying that the effect size of antidepressants compared to placebo is 8% actually means nothing. If you mean that the effect size is 0.08, then you are wrong. The actual effect size for response rate between antidepressants (59%, SD = 0.07) and placebo (46%, SD = 0.08) is 0.65, by my calculations.

That actually seems pretty high, and so I may be miscalculating. Can someone help me out?

This brings to mind the paper published PLoS in February which generated a bit of an uproar with its conclusion that SSRIs hardly help at all.

God, I hate doing this, but just one more comment on the above Khan study. When you factor in the initial response rates and include the subsequent lack of relapse, what the actual numbers show is interesting.

They show that of all the subjects who took antidepressants, 47% of them both responded and did not relapse for up to a year. The number is 27% for the placebo group. In other words, almost double the number of people who took antidepressants got well and stayed well up to a year compared to placebo. So, antidepressants came across as superior to placebo in their study.

However, the main notable conclusion of that study was that the placebo effect is SUSTAINED over time and is not transient. That, in and of itself, is very significant.

Take care.

There is a very similar sounding editorial in the British Journal of Psychiatry this month http://bjp.rcpsych.org/cgi/content/abstract/194/1/1
which also questions the effectiveness of randomised trials

so from what i can gather here, asking your child "how do you feel today" and really listening to the answer, engaging with your depressed child, is just about as effective as knocking them out with psych drugs.

and um, there are no withdrawal or side effects.

hmm.

in the 2+ years i've been reading about psych drug effectiveness and trials, a strikingly consistent impression i get is how effective placebo is, and not only relative to drugs.

i understand that conventional medicine has a problem prescribing placebo because the informed consent ethos would take away most of its power. so physicians have to stick to drugs. but in children, informed consent is with parents who could chose to give their kids placebos.

consider the ethics of the following scenario. a mother and 8-year-old consult a physician about bed wetting. the doctor, with all doctorly gravitas, issues a prescription, covertly mentioning to the mother that it's a sugar pill.

i'd say, if it works, great! if not, nothing lost.

i can understand why pharma wouldn't like such an idea but i wonder why physicians aren't thrilled to see such high placebo responses in depressed children.

tom:

Placebos essentially work through the conscious expectation of benefit and the unconscious prior conditioning within a therapeutic context. In other words, without the expectation that one will improve by taking a pill, which often comes from prior experience where one improves following medical treatment, then the placebo will be robbed of all its effectiveness.

I guess the problem with what you propose is that it presupposes a massive conspiracy of silence that ultimately will be exposed, as all conspiracies are -- thank God.

However, there could be a way of giving a patient a placebo without involving any inherent deception. A physician can say, "I can prescribe you a pill that studies have shown will help improve your condition. We don't have a good understanding of why it works, but know that in many people, it does. Are you willing to try it?"

There is nothing false in those statements, but the assumption by the patient is that the pill is biologically active, and if the truth came out that they are just placebos, then the entire placebo effect will be substantially reduced.

As Abraham Lincoln said, "You cannot fool all the people all of the time". Thank God for that, as well. :)

dguller: another way to give a placebo is homeopathy.

btw: i wasn't proposing anything in particular. i'm just fascinated by all the evidence in psych drug trials that points to its effectiveness.

and i'm fascinated by how the industry deals with this. on the one hand we have evidence-based medicine with researchers striving to find anything that works significantly better than a placebo. this is the camp from which come the uncomfortable meta studies about antidepressants. then on the other hand are the many physicians who dismiss that saying that they and their parents know from experience that the drugs work. they are the anecdotalists versus the statisticians.

the anecdotalists' argument, in my view, is the same as that of homeopathists and acupuncturists. to me, that seems not to fit so easily with their preference to prescribe other than placebo given the evidence that belief is a big part of whether or not a therapy works.

to use your words, expectation combined with prior conditioning can produce good therapeutic results. but the expectation involves an assumption that only a biologically active pill will produce the results. the evidence seems to show that this assumption is false. perhaps that could be corrected...

tom:

First, the evidence does show that antidepressants are superior to placebo with small to moderate effect sizes, especially for severe forms of depression. There is no battle between statisticians and anecdotalists in this issue. Nobody denies that they have been shown to work in significant ways, but whether their benefit outweighs the risks of adverse effects. That is the grey area here.

Second, you are correct that different expectations will lead to different placebo effects. The expectation that a witch doctor can banish evil spirits from your soul can result in a placebo effect during a religious ceremony. It is thought that that was how religious healing got its start thousands of years ago, i.e. piggy-backing on the placebo effect. So, perhaps in a new culture that does not have expectations of benefit from biological agents, then different placebo effects will result.

Take care.

Remember that one study does not set the rule. Properly executed research can only support or fail to support a hypothesis. The medical community won't have a theory to work with here until multiple studies with kosher methodology show similar results and future research starts giving an indication of the causation of what we're observing. So, the placebo effect data here are interesting, but not enough for us to all change course on accepted treatment methodologies.

It is, however, interesting to note that SSRI antidepressants have had an efficacy rate ony slightly above 50% in many adult studies. In May 2007, Harper's magazine ran an interesting cover story titled "Manufacturing Depression," which makes a well-researched and cogent argument against the ways in which we diagnose and treat depression. The author writes extensively about one particular medication trial in which he took part.

The "effect size" issue: In my opinion, I believe it is fine to say that the "effect size" of an ssri is the portion who improved, beyond the portion who improved with placebo. This is because, in my opinion, 1. there is not actually one single metric called "effect size" and 2. the dawdy effect size is exactly what we intuitively want to discover: what portion of some infinite, hypothetical population of patients will get better with treatment x versus treatment y? In other words, it is a relative indicator of effectiveness. "Eight percent more" is a totally valid, acceptable, understandable answer.

The sad aspect of all of this is that many of these efficacy studies do not sufficiently present data in order to allow us consumers of the data to figure out such things. Why don't each of these article report number-needed-to-treat, or Cohen's d, or Glass' delta? Why don't they consistently report means and standard deviations pre and post so we can figure this out, since authors and or journal reviewers/editors won't strive to include this basic info? Hmmm.