Talk:The Emperor's New Drugs

Latest comment: 9 years ago by Formerly 98 in topic "Most" bad results remain unpublished


"Powerful Placebo?"

edit

Would this term not be an oxymoron? I think what is meant is a placebo with powerful side effects -- could the term "Powerful placebo" be eliminated or modified?--Jrm2007 (talk) 10:56, 24 February 2014 (UTC)Reply

I think it's reasonable to describe a placebo that regularly produces a more powerful response than a dummy pill as a powerful placebo. I did a Google Scholar search for "powerful placebo" and "-Beecher" (Beecher wrote an influential and much-cited article in the 50s called "The Powerful Placebo") and found a number of instances of its use - mostly in relation to acupuncture and psychoanalysis. --Anthonyhcole (talk · contribs · email) 18:57, 24 February 2014 (UTC)Reply
The power of a placebo, is the power of belief, as in a religion. This power comes from the subconscious mind. An oxymoron is a combination of words that contradict each other. "jumbo shrimp" is an oxymoron. The placebo effect is real, so the two words "Powerful Placebo" do not contradict each other, it is not an oxymoron. http://www.cbc.ca/news/technology/pills-labelled-placebo-helped-mds-find-1.899620 Mark v1.0 (talk) 17:26, 4 October 2014 (UTC)Reply

Edits of Feb 1, 2015

edit

Hi @Anthonyhcole:,

Thanks for your note. This article is interesting to me and in many ways presents issues that are representative of many popular books written (both by experts such as Kirsch and by non-experts as well) on medical topics.

The book per se is of course Kirsch's views untrammeled by any sort of peer review, and on a controversial subject. There is a significant body of studies that disagree with his position on antidepressants in mild to moderate depression, and a large body (a majority I believe) that disagree with his position on antidepressants in severe depression.

You are entirely correct that the article I cited does not directly comment on or criticize Kirsch's results. But in discussing Kirsch's book, we inevitably discuss his views, which are controversial, and importantly, on a medical topic. I'm always very concerned that we present medical information and opinions in the most balanced way possible, as many unfortunately come here and use Wikipedia as the basis for medical decisions. However, few who oppose Kirsch's views present their own results explicitly as a rebuttal to Kirsch. To some degree it seems that we have a contradiction here between the requirements of WP:NPOV and the admittedly WP:OR-flavored nature of my edit. Its hard for me to see how we can fulfill both requirements simultaneously and provide our readers with an appropriately balanced view of the subject matter discussed in Kirsch's book.

I would appreciate your thoughts on how we can best handle this. This article should of course not present Kirsch's views as uncontested facts in Wikipedia's voice. But it seems to me the issues here are complex.

thanks, Formerly 98 (talk) 18:05, 2 February 2015 (UTC)Reply

"Most" bad results remain unpublished

edit

@Anthonyhcole: We currently have the sentence in the article

While analyzing antidepressant trials as part of his research into the placebo effect, Kirsch concluded that drug companies do not publish most of their disappointing drug trial results and do publish most favorable results, and that most decisions about the efficacy of a drug are based only on published results."

We've gotten into a disagreement on the whether to attribute this statement to Kirsch's or state it in Wikipedia's voice as "established fact". But at this point we don't even have a source that Kirch ever stated that "most negative studies remain unpublished", let alone that it is "established fact". Can we get source for this statement? Otherwise I'm inclined to go beyond simply attributing this statement to Kirsch, and think it should be deleted altogether.

Ditto for the sentence:

"most decisions about the efficacy of a drug are based only on published results..."

In fact the FDA and EMA both consider unpublished results in determining whether a drug meets the statutory requirements for safety and efficacy. So not only is this phrase unsourced, it appears to be demonstrably incorrect. Please find a source if you want to keep it in the article. Formerly 98 talk|contribs|COI statement 22:21, 30 March 2015 (UTC)Reply

I removed the statement "Most decisions about drugs are made based on published results" per the information above. Also modified text to make it clear that Kirsch was describing antidepressants. Whether the phenomenon is general to other classes of drugs or not is not the issue here; Kirsch's book describes antidepressants and not the entire range of pharmaceuticals across therapeutic areas. We should not attribute statements to Kirsch unless he actually made them. Formerly 98 talk|contribs|COI statement 08:50, 31 March 2015 (UTC)Reply
In terms of unpublished negative results on antidepressants, the best source is probably Turner et al., NEJM 2008. This study found that of 36 "negative" clinical trials of antidepressants registered with FDA, 22 were not published at all, and 11 were published but in a way that suggested the results were positive. Only 3 were presented as truly "negative" results in publication. So it is technically correct to say that "most" (>50%) negative studies of antidepressants are never published. Formerly98 is correct that these results are nonetheless presented to FDA, which is how the Turner et al. study was able to obtain them for comparison to the published literature.

As to "most decisions"... I think Formerly98 is correct that regulatory approval by FDA is based on review of all clinical trials, including unpublished ones. However, the clinical decision to use an antidepressant is often based on clinicians' review of the published literature. Likewise, meta-analyses and systematic reviews of the efficacy of antidepressants may fail to account for these negative unpublished trials (some of which are available only through FOIA requests), and thus overstate the efficacy of antidepressants. In this sense, it is correct to say that most clinical decisions about the efficacy of a drug, and whether to use it in a specific patient or population, are based on the published literature. MastCell Talk 15:58, 31 March 2015 (UTC)Reply

Hi MastCell:
I think you make some great points. However, every meta analysis that I've seen over the last few years explicitly evaluates the literature for publication bias, and most of the better ones include unpublished results taken from FDA FOI Act requests. So I'll summarize as follows:
  • Did Kirsch actually make this statement in his book or is this WP:SYNTHESIS? I think it is fair to ask for a page number or a quote. If Kirsch did not say it, its just WP:COATRACK
  • If Kirsch did say it, I think its fine to include, but we should qualify exactly what he meant. If he meant post approval decisions on clinical guidelines, that is realistic. If he meant approval decisions, I have no idea what he is talking about.
On a broader but somewhat off-topic note, the FDA has a very different viewpoint on this than does Kirsch. Kirsch and many other in investigators critical of antidepressants treat trials in which both the test drug and a positive control fail to do better than placebo as evidence against the efficacy of antidepressants. The FDA considers these "failed experiments" and does not incorporate the results of such trials in its global efficacy analysis. While there are good arguments on both sides of that debate, I tend to agree with FDA. The "null hypothesis" being tested as the primary efficacy endpoint in a typical antidepressant trial is "drug is no better than placebo". The goal of the trial is to disprove this null hypothesis by showing a difference in mean changes from baseline that has only a 5% chance of occuring by chance. Failure to reach statistical significance is not evidence that the drug does not work, it is failure to prove the null hypothesis is false. More pithily, "absence of evidence is not evidence of absence". Obviously this analysis breaks down if you do 100 trials and only hit stat signficance in two of them, but I am not aware of any case like that.
Looking forward to hearing your proposal for how to best move forward. Best Formerly 98 talk|contribs|COI statement 16:15, 31 March 2015 (UTC)Reply
Regarding this: it's nearly two years since I read the book but my memory is that Kirsch reported that he "realised" not all trial results were published - that is, it was news to him when he first approached the project. MastCell's analysis is correct: I was referring to clinical decisions (as well as recommendations in guidelines, conclusions in reviews, etc.). I'll get the book out of the library again and find some page numbers. --Anthonyhcole (talk · contribs · email) 16:20, 1 April 2015 (UTC)Reply

Thank you, that sounds like a great way to resolve this. Formerly 98 talk|contribs|COI statement 16:36, 1 April 2015 (UTC)Reply