The burden of proof lies on yo who makes the positive claim. "X works" is a positive claim. Of course, it's more complicated than that in practice, but it's *that* extra context that is often lost in translation or of questionable applicability when we talk about "everyone".
alex: There was study in the NEJM that appears to me to use the exact same data set, but makes much more cautious conclusions. The PLOS article is sensationalist
SisterY: Acupuncture can be effective for controlling pain. Mechanism is unclear, but I believe there are trials already showing it works.
The study is based on change in a numeric scale for depression. If I had to make a guess, I would say the measuring instrument they are using is just not sensitive for important changes in peoples mood/behavior, rather than SSRIs don't work. The drugs themselves have probably undergone more rigorous study than the measuring instrument. Also, the drugs have also probably undergone more thorough study than any alternative treatment.
"Why is "Overcoming Bias" perpetuating the pervasive anti-drug bias that keeps so many people from seeking treatment?"
Forgive me, I'm not very intelligent. Robin & others here present actual evidence that SSRIs don't work for mild-to-moderate depression (the cases for which they are most often prescribed) and you describe this as anti-drug bias? Are you arguing a pro-drug bias is better? That we should take drugs even when they don't work?
Robin is not arguing against treating dysthymia or moderate depression - the discussion might soon shift to what kind of treatments have been proven to work. The problem is in medicine, which is supposed to be scientific, society often pays through the nose for that which actually doesn't have adequate scientific evidence behind its efficacy.
This is my 3rd comment, so I'm outta here. But please instead of so many chiming in with "I'm a happy placebo lover who believes in my impressive shaman-psychiatrist," I'd be interested in seeing what has been proven to work for dysthymia, and if nothing has been studied, I'd like to hear commentators tell me why not.
Why is "Overcoming Bias" perpetuating the pervasive anti-drug bias that keeps so many people from seeking treatment? The very same skepticism and fear of stigma kept me from going in for treatment for a very long time. By the time I did, things were not going well in my professional life, and I'd tried and tried to "cheer up" or "snap out of it" on my own. Now I wish I'd gone in for help years earlier. Maybe it is a placebo effect; but I was astonished to find that I felt better than I had in years. I feel like wind is at my back. I am able to think about the problems in my life without getting sucked into a stomach-wrenching depression, and I even feel excited about facing the challenge and growing through it. My mood and perspective has changed in ways I couldn't have imagined. Now I have the emotional energy to keep my life together, eat healthy, and take care of myself even while I face some real challenges ahead. I've had friends who took sertraline and saw similar transformations in their lives.
I have a friend who has been suicidal on and off for months now, and I can't convince her to go get help, precisely because of all this anti-drug bias and stigma. And the very fact that sertraline can cause manic episodes if you overdose or if you've got misdiagnosed manic-depression, tells me that there's something real going on. Even if it IS a placebo effect, let's figure out how to harness that power without the side-effects.
No, my parents' insurance company pays lots of money to Big Pharma. I don't pay a cent. I went on my first antidepressant in high school when my parents demanded it; I didn't want to take it, insisting that there was nothing wrong with me, the problem was with the rest of the world. (What I wanted to do was stop going to school. My parents wouldn't let me, referring to a family friend who got summoned to family court because his daughter was late to school every day.)
It might just be post hoc fallacy, but my attempts to go off medicine have always been followed fairly quickly by trouble (or, at least, what my parents call trouble). The only thing that I can identify as having any effect on me at all during my sophomore year of high school was the pills I was given. After all, life still sucked, but I now I was cheerful about it!
"things like tremors, abnormal sleepiness, and loss of sexual desire are mild."
Clearly you've never talked to the wife of a guy on an SSRI long-term - now that's depression. Dropping coffee cups everywhere and nodding off on the job or while driving really aren't a party. But may I gently say it seems condescending of you to decide on your own authority that these issues are "mild" and should just be endured. Yikes!
@Doug S
"an effect that helps me somehow and they have some nasty withdrawal symptoms."
Let me politely ask: are you saying you pay lots and lots of money to Big Pharma every year for an addictive substance with "nasty withdrawal symptoms" for an effect you can't really describe? Does this mean you keep taking these expensive medications with unclear results simply to avoid the pain of withdrawal? I don't mean to criticize, because I am sure you know what is best for you, but doesn't something seem strange in this scenario?
Stuff like this moves me more and more towards Robin's counter-intuitive "we have too much medicine" position.
SSRIs definitely do something. Whether that something is better than placebo at alleviating symptoms of depression remains an open question, but I keep taking my pills because they seem to have an effect that helps me somehow and they have some nasty withdrawal symptoms.
(Technically, I'm taking an SNRI rather than an SSRI, but they're similar enough.)
There have been quite a few independent studies showing that antidepressant drugs are only barely better than placebos. Even if this particular study has holes knocked in it, you'd need to account for the entirety of the available data.
I'm really puzzled by their claim that statistically and clinically significant improvements were due to a "decrease in the response to placebo rather than an increase in the response to medication." What else could "responding to a drug" possibly mean, other than a difference compared to the placebo group?! The placebo doesn't go up or down; it doesn't go anywhere; it's your baseline. You don't measure it; it's what you use to measure other things. It's like saying, "I didn't gain weight, the kilogram got lighter!"
In summary: nope, that wasn't in my statistics class.
consider it as a good Bayesian would: how much should your posterior probability change given this one study and the large number of priors?
It is technically incorrect to write, "the large number of priors." You mean, "the high prior probability" (that antidepressants work, at least partially, at least for some depressions). There is only one prior in a Bayesian update; there is only one prior in Bayes' theorem. Since Bayesian updating is such a central concept for this blog, I decided to offer a correction even though everyone probably knows what you meant.
Um. Robin, first of all, this is one meta-analytic study of questionable methodology, producing a weakly positive result, and referring not to SSRI data as a whole but to four new antidepressants, only two of which are SSRIs, during clinical trials. It produced a result which is very strongly contradicted by both clinical trials and extensive clinical experience over the last twenty-odd years. So it's inappropriate to draw any very strong conclusions; consider it as a good Bayesian would: how much should your posterior probability change given this one study and the large number of priors?
As to your question to Joseph, you're erecting a straw man there. The issue is not whether you ought act on a study if any relevant information was missing; the question is whether you ought act on this study, when extremely central information is missing or explicitly wrong.
There's a further question whether one ought to draw conclusions about all antidepressants ("Antidepressants fail") given that this only examines a selected data set for a small number of antidepressants. On that basis alone, your initial assertion and headline are mistaken and should be corrected.
As far as your other, rather risible, clinical advice --- "Given their severe side-effects, I'm not sure I'd advise anti-depressants even when other treatments have been ineffective" --- I'm tempted crankily simply to note that I wasn't aware that you had an MD degree. That's not really sufficient, however, considering how supremely ignorant that statement truly is. First, SSRIs in general have rather mild side effects: in most people, most of the time, the side effects are a lot less troubling than the depression. Frankly, in a disease as generally debilitating as depression, things like tremors, abnormal sleepiness, and loss of sexual desire are mild. The mortality rate from extended chronic depression is comparable to some lesser cancers; a chemotheraputic agent for low-mortality cancer that had a similar side-effect profile to the SSRIs would be considered a godsend.
The notion that SSRIs lead to an increase in suicidal thoughts and actions is somewhat hard to identify clinically, although it has gotten a lot of anecdotal play in the press; what's usually forgotten is that people moving from an acute depressive episode to remission are inherently more prone to suicide than the general population even if there has been no treatment at all, and young people with depression are apparently more prone to suicide than older people. Thus, even if SSRIs had no side effects whatsoever, and only caused gradual remission from an acute depressive episode, you would still expect an elevated suicide rate, and especially so in the young.
Robin,Your quantifiers are a ridiculous interpretation of Joseph's comment. This is a meta-analysis, which should be held to higher standards of inclusivity and it is a large class of data that they ignored. When a meta-analysis drops data, it is a warning of over-fitting.
On the other hand, the study is explicit that its choice of data is to avoid publication bias.
I'd say that's because my brain is underproducing (or over uptaking; whichever) serotonin, thus it brings me up to more normal brain function Unlikely - the brain is extremely good at adapting to different levels of neurotransmitters.
Perhaps you have a defective capacity to compensate, or perhaps the drug is causing parts of your brain to react in ways they wouldn't normally, but it is extraordinarily implausible for your claim to be true on its own grounds.
The burden of proof lies on yo who makes the positive claim. "X works" is a positive claim. Of course, it's more complicated than that in practice, but it's *that* extra context that is often lost in translation or of questionable applicability when we talk about "everyone".
methodological flaws in the study are reviewed at log base 2
alex: There was study in the NEJM that appears to me to use the exact same data set, but makes much more cautious conclusions. The PLOS article is sensationalist
SisterY: Acupuncture can be effective for controlling pain. Mechanism is unclear, but I believe there are trials already showing it works.
The study is based on change in a numeric scale for depression. If I had to make a guess, I would say the measuring instrument they are using is just not sensitive for important changes in peoples mood/behavior, rather than SSRIs don't work. The drugs themselves have probably undergone more rigorous study than the measuring instrument. Also, the drugs have also probably undergone more thorough study than any alternative treatment.
@wellbutrin
"Why is "Overcoming Bias" perpetuating the pervasive anti-drug bias that keeps so many people from seeking treatment?"
Forgive me, I'm not very intelligent. Robin & others here present actual evidence that SSRIs don't work for mild-to-moderate depression (the cases for which they are most often prescribed) and you describe this as anti-drug bias? Are you arguing a pro-drug bias is better? That we should take drugs even when they don't work?
Robin is not arguing against treating dysthymia or moderate depression - the discussion might soon shift to what kind of treatments have been proven to work. The problem is in medicine, which is supposed to be scientific, society often pays through the nose for that which actually doesn't have adequate scientific evidence behind its efficacy.
This is my 3rd comment, so I'm outta here. But please instead of so many chiming in with "I'm a happy placebo lover who believes in my impressive shaman-psychiatrist," I'd be interested in seeing what has been proven to work for dysthymia, and if nothing has been studied, I'd like to hear commentators tell me why not.
Why is "Overcoming Bias" perpetuating the pervasive anti-drug bias that keeps so many people from seeking treatment? The very same skepticism and fear of stigma kept me from going in for treatment for a very long time. By the time I did, things were not going well in my professional life, and I'd tried and tried to "cheer up" or "snap out of it" on my own. Now I wish I'd gone in for help years earlier. Maybe it is a placebo effect; but I was astonished to find that I felt better than I had in years. I feel like wind is at my back. I am able to think about the problems in my life without getting sucked into a stomach-wrenching depression, and I even feel excited about facing the challenge and growing through it. My mood and perspective has changed in ways I couldn't have imagined. Now I have the emotional energy to keep my life together, eat healthy, and take care of myself even while I face some real challenges ahead. I've had friends who took sertraline and saw similar transformations in their lives.
I have a friend who has been suicidal on and off for months now, and I can't convince her to go get help, precisely because of all this anti-drug bias and stigma. And the very fact that sertraline can cause manic episodes if you overdose or if you've got misdiagnosed manic-depression, tells me that there's something real going on. Even if it IS a placebo effect, let's figure out how to harness that power without the side-effects.
No, my parents' insurance company pays lots of money to Big Pharma. I don't pay a cent. I went on my first antidepressant in high school when my parents demanded it; I didn't want to take it, insisting that there was nothing wrong with me, the problem was with the rest of the world. (What I wanted to do was stop going to school. My parents wouldn't let me, referring to a family friend who got summoned to family court because his daughter was late to school every day.)
It might just be post hoc fallacy, but my attempts to go off medicine have always been followed fairly quickly by trouble (or, at least, what my parents call trouble). The only thing that I can identify as having any effect on me at all during my sophomore year of high school was the pills I was given. After all, life still sucked, but I now I was cheerful about it!
@Charlie
"things like tremors, abnormal sleepiness, and loss of sexual desire are mild."
Clearly you've never talked to the wife of a guy on an SSRI long-term - now that's depression. Dropping coffee cups everywhere and nodding off on the job or while driving really aren't a party. But may I gently say it seems condescending of you to decide on your own authority that these issues are "mild" and should just be endured. Yikes!
@Doug S
"an effect that helps me somehow and they have some nasty withdrawal symptoms."
Let me politely ask: are you saying you pay lots and lots of money to Big Pharma every year for an addictive substance with "nasty withdrawal symptoms" for an effect you can't really describe? Does this mean you keep taking these expensive medications with unclear results simply to avoid the pain of withdrawal? I don't mean to criticize, because I am sure you know what is best for you, but doesn't something seem strange in this scenario?
Stuff like this moves me more and more towards Robin's counter-intuitive "we have too much medicine" position.
SSRIs definitely do something. Whether that something is better than placebo at alleviating symptoms of depression remains an open question, but I keep taking my pills because they seem to have an effect that helps me somehow and they have some nasty withdrawal symptoms.
(Technically, I'm taking an SNRI rather than an SSRI, but they're similar enough.)
There have been quite a few independent studies showing that antidepressant drugs are only barely better than placebos. Even if this particular study has holes knocked in it, you'd need to account for the entirety of the available data.
A lot of critical commentary about the study is at this blog.
I'm really puzzled by their claim that statistically and clinically significant improvements were due to a "decrease in the response to placebo rather than an increase in the response to medication." What else could "responding to a drug" possibly mean, other than a difference compared to the placebo group?! The placebo doesn't go up or down; it doesn't go anywhere; it's your baseline. You don't measure it; it's what you use to measure other things. It's like saying, "I didn't gain weight, the kilogram got lighter!"
In summary: nope, that wasn't in my statistics class.
consider it as a good Bayesian would: how much should your posterior probability change given this one study and the large number of priors?
It is technically incorrect to write, "the large number of priors." You mean, "the high prior probability" (that antidepressants work, at least partially, at least for some depressions). There is only one prior in a Bayesian update; there is only one prior in Bayes' theorem. Since Bayesian updating is such a central concept for this blog, I decided to offer a correction even though everyone probably knows what you meant.
Um. Robin, first of all, this is one meta-analytic study of questionable methodology, producing a weakly positive result, and referring not to SSRI data as a whole but to four new antidepressants, only two of which are SSRIs, during clinical trials. It produced a result which is very strongly contradicted by both clinical trials and extensive clinical experience over the last twenty-odd years. So it's inappropriate to draw any very strong conclusions; consider it as a good Bayesian would: how much should your posterior probability change given this one study and the large number of priors?
As to your question to Joseph, you're erecting a straw man there. The issue is not whether you ought act on a study if any relevant information was missing; the question is whether you ought act on this study, when extremely central information is missing or explicitly wrong.
There's a further question whether one ought to draw conclusions about all antidepressants ("Antidepressants fail") given that this only examines a selected data set for a small number of antidepressants. On that basis alone, your initial assertion and headline are mistaken and should be corrected.
As far as your other, rather risible, clinical advice --- "Given their severe side-effects, I'm not sure I'd advise anti-depressants even when other treatments have been ineffective" --- I'm tempted crankily simply to note that I wasn't aware that you had an MD degree. That's not really sufficient, however, considering how supremely ignorant that statement truly is. First, SSRIs in general have rather mild side effects: in most people, most of the time, the side effects are a lot less troubling than the depression. Frankly, in a disease as generally debilitating as depression, things like tremors, abnormal sleepiness, and loss of sexual desire are mild. The mortality rate from extended chronic depression is comparable to some lesser cancers; a chemotheraputic agent for low-mortality cancer that had a similar side-effect profile to the SSRIs would be considered a godsend.
The notion that SSRIs lead to an increase in suicidal thoughts and actions is somewhat hard to identify clinically, although it has gotten a lot of anecdotal play in the press; what's usually forgotten is that people moving from an acute depressive episode to remission are inherently more prone to suicide than the general population even if there has been no treatment at all, and young people with depression are apparently more prone to suicide than older people. Thus, even if SSRIs had no side effects whatsoever, and only caused gradual remission from an acute depressive episode, you would still expect an elevated suicide rate, and especially so in the young.
Robin,Your quantifiers are a ridiculous interpretation of Joseph's comment. This is a meta-analysis, which should be held to higher standards of inclusivity and it is a large class of data that they ignored. When a meta-analysis drops data, it is a warning of over-fitting.
On the other hand, the study is explicit that its choice of data is to avoid publication bias.
I'd say that's because my brain is underproducing (or over uptaking; whichever) serotonin, thus it brings me up to more normal brain function Unlikely - the brain is extremely good at adapting to different levels of neurotransmitters.
Perhaps you have a defective capacity to compensate, or perhaps the drug is causing parts of your brain to react in ways they wouldn't normally, but it is extraordinarily implausible for your claim to be true on its own grounds.
Sertraline has worked wonders in my life. I've been taking it for nearly 10 years now. I wish I had found it 20 years earlier. Life is good.