The authors of a study allegedly showing that antidepressants prevent suicide have retracted it over unspecified errors. Here’s the notice:
At the request of the authors and in agreement with the Editor-in-Chief and Wiley-Blackwell, the following article “Antidepressant medication prevents suicide in depression”. Acta Psychiatr Scand 2010;122:454–460 has been retracted. The retraction has been requested and agreed due to unintentional errors in the analysis of the data presented.
The original paper linked “data on the toxicological detection of antidepressants in 18 922 suicides in Sweden 1992–2003” to “registers of psychiatric hospitalization as well as registers with sociodemographic data.” It found:
The probability for the toxicological detection of an antidepressant was lowest in the non-suicide controls, higher in suicides, and even higher in suicides that had been psychiatric in-patients but excluding those who had been in-patients for the treatment of depression.
Its conclusion was a bit convoluted:
The finding that in-patient care for depression did not increase the probability of the detection of antidepressants in suicides is difficult to explain other than by the assumption that a substantial number of depressed individuals were saved from suicide by postdischarge treatment with antidepressant medication.
The paper has been cited just four times, according to Thomson Scientific’s Web of Knowledge. But we understand that other studies by lead author Goran Isacsson are often used by those who support the idea that antidepressants cut down on suicides, and as evidence that the FDA’s “black box warning” on antidepressants should be rewritten. That warning reads, in part:
Antidepressant medicines may increase suicidal thoughts or actions in some children, teenagers, and young adults when the medicine is first started.
Isacsson and a co-author debated the issue with Melissa Raven and a co-author in the June 2010 issue of the British Journal of Psychiatry. At one point, Raven and Jon Jureidini conclude:
In summary, Isacsson & Rich’s premises are flawed and they have overstated their case with selective citation and biased interpretation of evidence.
We’ve contacted Isacsson as well as the editor of Acta Psychiatrica Scandinavica to find out what the “unintentional errors” were, and whether any other papers will be affected, and will update with anything we find out.
Update, 11:45 a.m. Eastern, 3/19/12: Isacsson tells us:
We discovered lately that there was a coding error regarding diagnoses in the database we utilized for the 2010 paper. When corrected, antidepressants were detected in depressed suicides as often as could be expected and not less than expected which was the crucial finding in the paper. This means that no conclusion can be drawn from the study regarding antidepressants’ effects on suicide risk in any direction.
The database has not been used for other studies so no other papers are affected.
Hat tip: Bernard Carroll
That retraction is almost entirely uninformative (but sadly typical as RW readers know) – we need to get the details here.
It’s notable that many of Isaacson’s papers on this topic were published in the same journal Acta Psychiatr Scand.
I have no knowledge of this particular paper or what might be wrong with it, so please don’t misconstrue my comments….
A colleague sat on the committee that ended up requiring the “black box” warning on antidepressants used for adolescents. It was a highly-political and deeply-contentious process and my colleague wound up feeling that the “black box” warning was a mistake. The data set they evaluated had no completed suicides at all, even though an enormous number of patients were involved, so the warning is based on reported thoughts that were never acted upon successfully.
Psychiatrists have known for years that a deeply-depressed person who begins antidepressants is more likely to commit suicide in those first weeks after beginning medication. This may be because patients suddenly have energy enough to act upon the thoughts they’ve been having all along.
I worry that this retraction will tip the balance further away from what I believe to be appropriate use of antidepressant medication.
FYI, Acta Psychiatr. Scand. is thought to be a highly reputable journal…..
You are certainly right that “psychiatrists have known for years…” and it makes sense that patients would feel more energized and more likely to carry out plans, tho still depressed, after a week or two on an antidepressant. That is an impression based on clinical experience and I don’t know if it could be confirmed by research. So much of psychiatric “knowledge” is clinical in this way and it would take volumes of research to verify it all.
The same is true of the black box warning– it is not really buttressed by good research and appears to be anecdotal based on a few case studies. I don’t think that is a good way to decide on a black box warning, considering how much the FDA tries to rely on good research.
We can only continue, for now, with the clinical impressions we have, particularly the perception that antidepressants are better treatment than just talk for severely depressed patients. That perception, I’m afraid, is also not backed by good evidence, no matter how reasonable it seems. If I’m wrong, please correct me, but I have gotten the impression that evidence of efficacy for many antidepressants is extremely weak.
It appears that Isaacson is convinced that antidepressants are helpful, which makes his research open to question due to lack of neutrality. He had to retract this study and the others will be undergoing more scrutiny. So we’re still back on square one. No wonder you can never get two doctors to agree on a treatment plan.
Evidence for the efficacy of antidepressants is indeed weak, but it does exist. My favorite paper on this is Walsh et al (JAMA. 2002;287(14):1840-1847), which found that average efficacy of antidepressants across 75 RCTS was 50.1%. This is really not much better than the average efficacy of placebo in those same 75 trials, which was 29.7%. Depression waxes and wanes over a few weeks and most antidepressant RCTs last only a few weeks, so we are in the position of having equivocal evidence.
Nevertheless, it seems to me that the efficacy of antidepressants is far better demonstrated than the efficacy of the “talking cure.” How can you have a placebo for the talking cure? And without a placebo, blinded trials are impossible and all of the evidence may be a result of observer bias.
I note that every researcher who has ever done a study is convinced that the treatment they study could be helpful. If optimism (or lack of neutrality) is equivalent to bias, then we all have it….
Isacsson wrote me after I questioned him about the retraction.
“We discovered miscoded diagnoses in the database. When this was corrected were antidepressants found in the expected range. Thus, our conclusion cannot be drawn, neither any contrary conclusions. The database was created for this paper and the error has not affected any other of our studies.
With kind regards”
Göran Isacsson
Despite his reassurances, his other studies will undergo more critical scrutiny.
To the esteemed Steen:
I’ll take 50% over 30% if the sample size is big enough–especially if the timing is right (you tell them two weeks, and it is two weeks–then they have more reason to trust you.)
It’s true, you’d have to have some investment in a treatment just to study it in the first place; but sometimes an unhealthy attachment can occur, you know… and now, if I could just come up with the right wording for my prayer, God will listen to me this time…
The study was accepted for publication March 4, 2010. Isn’t it remarkable that TWO years later one of the co-authors discovers that the most important piece of data in the study is wrong?
Isacsson wrote in the article:
“The main finding is that the category of suicides in which the detection of antidepressants could be expected to be most frequent, i.e. the suicides that had been diagnosed with depression prior to death, instead constituted the diagnostic category of hospitalized suicides that had the lowest frequency of detection of antidepressants (15.2%).”
Everything in the article and its conclusion about “prevented suicides” is built on this low “observed percentage”, 15,2% (164 persons).
But this percentage was wildly wrong. Facts are that 50% (!!) in this group (of 1077 persons) hade received antidepressants. Meaning we can increase the original group with 230% (!) Meaning they built the article and its findings on 164 persons, when the correct number was around 540 – they missed around 375 persons.
Hard to get much out of the retraction, but I’m concerned readers will draw conclusions that the whole body of research finding that antidepressants do NOT in fact cause suicidal behaviors (or suicides, as the publi seems to think) will be suspect. Robert Gibbons from the U of C raised serious doubts about the FDA’s link as early as 2007, with his “Early Evidence on the Effects of Regulators’ Suicidality Warnings on SSRI Prescriptions and Suicide in Children and Adolescents” in the American Journal of Psychiatry; as early as 2006 (as the FDA warnings were still going on), Simon found in his “Suicide Risk During Antidepressant Treatment” in the American Journal of Psychiatry that suicidality declined in the months after starting an antidepressant when compared with the month previous to initiation, and a recent Gibbons study, “Suicidal thoughts and behavior with antidepressant treatment: reanalysis of the randomized placebo-controlled studies of fluoxetine and venlafaxine” from the Feb Archives of General Psychiatry disproved fairly significantly the connection of Prozac (one of the first to receive the FDA’s black-box warning) with suicidaal behavior and suicides. [Can see full post on the topic at http://wp.me/p22afJ-1sU.%5D It’s just important to remember that one retracted study does not al findings on the topic.
Psychiatrist Goran Isacsson withheld unwanted research data about antidepressants and suicide
The article “Antidepressant medication prevents suicide in depression” was retraced by the journal Acta Psychiatrica Scandinavica in March 2012. The reason for the retraction was said to be “unintentional errors in the analysis of the data presented”.
But what was wrong and what were the correct data in the study? Well, it took a five month legal process to force the author psychiatrist Goran Isacsson and Karolinska Institutet in Stockholm to release the correct data about antidepressants and suicide.
Read full article here: http://jannel.se/Isacsson.Destroy.Antidepressants.pdf
Janne, the ever-incisive 1 boring old man has done a blog about your article:
at its worst…
http://1boringoldman.com/index.php/2013/01/11/at-its-worst/
He also did a blog about the retraction in March last year:
there to do…
http://1boringoldman.com/index.php/2012/03/20/there-to-do/
The question to ask is was it an honest mistake or did that have something to do with the declared conflict of interests? Should be the people who study drugs be allowed to take pharma money at all? It seems radical to ask this question but to me it reeks of corruption.