Posted by Larry Hoover on May 25, 2009, at 10:00:55
In reply to Re: Antidepressants Hardly Help ( Time Magazine) » Larry Hoover, posted by 49er on May 17, 2009, at 6:17:41
> Hi Larry,
Hi, 49er. I'd like to address your closing remark first. You said:
> Anyway, I can't even compete with you in my dreams in analyzing data so if I am missing something that is relevant, please accept my apologies in advance.I don't mean to come across as intimidating. I really don't.
If I spend some time reviewing a study, I try to be as thorough as possible. If I end up writing about a study, I tend to include all the things I thought about. Some studies are so straight-forward, there's really not a lot to say other than "the evidence shows that....blah blah". In other cases, the study conclusions are non sequitur, i.e. they do not follow from the evidence. Or they are severely limited, i.e. not generalizable. Or whatever. It seems to me that the most contentious papers are also severely flawed.
I like to present scientific axioms because they summarize conservative scientific thought. One such axiom is, "Extraordinary claims require extraordinary evidence." So, if the lay press gets hold of an extraordinary claim, and distorts it (typically) to get that sound bite that grabs attention, and I find that the underlying evidence is not correct, let alone extraordinary, I'm all over it.
> Is the study really outrageous in light of this statement by Dr. Robert Heydaya, a psychopharmacologist who is definitely not anti meds? This is an exert from a response he posted to someone who had made a comment on his article.
>
> http://tinyurl.com/p5kaok
>
> "There are three issues I am raising. First, there is overreliance on antidepressants-action is taken without looking at the WHOLE picture, trying to find underlying correctable causes, which if treated would improve one's overall health and prevent other diseases. Second, depending on the study you read, only 1/3 to 1/2 of all patients treated with antidepressants have a full recovery with the current methods."If you're referring to the Kirsch paper as the one that might be outrageous, the answer is yes, it is outrageous. More in a moment.
I agree with what Dr. Heydaya says. The idea that simply taking a friggin' pill is going to fix depression is deeply ingrained in our culture, but it is very inappropriate. If you don't look at stressors and social supports and diet and sleep and exercise and comorbidities and so on, then what do you expect to obtain from an antidepressant? That these factors magically melt away? Somehow? How?
And yes, current treatments have limited success, as measured by short-term studies (some as few as 4 weeks), with all other variables controlled (diet, stressors, blah blah) other than social supports (you get a lot of attention in a clinical trial), with the threshold set at full recovery. Nonetheless, they are significantly better than placebo.
> Also, since a successful response is based on a 50% reduction in symptoms, that has to be accounted for in the overall success rate.
Absolutely. A partial response is still a response. It's a starting point to full recovery, one should hope.
> Additionally, the whole premise of pharma companies who are claiming that antipsychotics are effective add-ons for depression is that ADs only work 33% of the time. So if they are lying, then shouldn't the FTC be getting them for false advertising?
Nobody's lying. They're discussing an alternative to monotherapy. I still question the lack of attention to diet, exercise, etc., but that's not the drug companies' job. It's your doctors' job. And your job.
> Your point would make more sense if other folks were coming up with higher success rates. But that isn't what is happening.My point is about the distortion of evidence to fit an intellectual perspective which is demonstrably false. The Kirsch papers. I consider Kirsch to be intellectually dishonest. Here is a non-exhaustive summary of why I believe that to be the case.
Way back in 2002, Kirsch wrote the first paper I quote below. About six years later, he publishes the second one. Here's what he said about the data he examined:
"'The Emperor's New Drugs: An Analysis of Antidepressant Medication Data Submitted to the U.S. Food and Drug Administration'(2002)
We received information about 47 randomized placebo controlled short-term efficacy trials conducted for the six drugs in support of an approved indication of treatment of depression. The breakdown by efficacy trial was as follows: fluoxetine (5), paroxetine (16), sertraline (7), venlafaxine (6), nefadozone (8), and citalopram (5).""'Initial Severity and Antidepressant Benefits: A Meta-Analysis of Data Submitted to the Food and Drug Administration'(2008)
Forty-seven clinical trials were identified in the data obtained from the FDA. The dataset comprised 35 clinical trials (five of fluoxetine, six of venlafaxine, eight of nefazodone, and 16 of paroxetine) involving 5,133 patients, 3,292 of whom had been randomized to medication and 1,841 of whom had been randomized to placebo."Both studies, six years apart, analyze the exact same evidence, except in the latter one, he excludes data for two drugs. In the latter paper, he does not reveal that he had this information for six years. He makes it sound like he just obtained it, through Freedom of Information requests. My post immediately prior to this one included published and unpublished evidence with respect to the efficacy of 12 antidepressants. Why did Kirsch not reveal that he was re-analyzing data he'd held for six years, and why did he not obtain and analyze data available to others?
The second thing he did was that he retroactively applied an efficacy standard (the NICE guideline) to data collected for another purpose entirely. Notwithstanding the name applied to the studies (clinical trials), they are not intended to demonstrate or predict response outside of the trial environment, i.e. clinical efficacy. They are short-term efficacy trials, and nothing more than that. I know of only one clinical efficacy trial, the STAR*D trial. Although NICE discusses at great length the limitations inherent in applying their standard to simple efficacy trials, Kirsch never once goes there.
Here is the link to the NICE document, which covers every possible treatment for depression, and the evidence that supports (or fails to support) that treatment. I highly recommend that people consider this document as one of the best resources available to them, for evidence-based treatment of depression.
http://www.nice.org.uk/nicemedia/pdf/CG23fullguideline.pdfNow, back to Kirsch. Althought he lifted the NICE guideline from the previously published document (above), or its predecessor (2004), he totally fails to acknowledge that NICE found that antidepressants, and specifically the SSRIs, are recommended as first-line treatment for all but mild depression. And in mild depression that persists, they do recommend antidepressant treatment.
Kirsch then applies statistical treatments to the clinical trial results that are totally inappropriate. The HAM-D (Hamilton Depression Scale) is an ordinal scale. It's not like height in inches (an interval scale), where every inch is the same size, and if you measure it twice you should get the same result, or if someone else measures it, they should get the same result. Instead, for the HAM-D, the numerical value obtained is the sum of the individual's impression of their depressive symptoms. The interpretation of the questions and the answers is probably just as individual.
Just to contextualize that, I've seen people post about having the most severe depression imaginable. It's affecting their work, and wowzers, it even makes going to a party unenjoyable. Eh, what?!? You can work and go to parties? I've gone years unable to work, or lengths of time unable to feed myself and bathe regularly. Party? Would that other guy's score on the HAM-D be comparable to mine, and indicate similar degrees of depression?
Moreover, the test-retest reliability of the HAM-D is not perfect. If you give the test twice to the same person, but rearrange the questions, there is often a different score. Also, consider if the difference between a score of say, 5 and 6 (not depressed), is the same magnitude of difference as that between 25 and 26 (severely depressed)? They both change by one, but are the changes comparable? Who knows?
I hope you get where that takes us. Statistically, you are permitted to rank the scores, and to calculate difference scores for an individual (the change in score over time), but nothing more than that. You can graph how many changed how much in each of a drug or placebo group, or compare percentile ranks in distinct ranges of scores, but that's it.
It has become standard practise to calculate mean changes, and standard deviations, and the statistical significance of any differences in these values for the HAM-D, but strictly speaking, you shouldn't. But absolutely forbidden is what Kirsch did. He did standardization, linear regression, and quadratic factor analysis on statistics (the mean change scores) which themselves shouldn't have been determined. Not only that, he did multiple analyses, and selected the one with the worst outcome for the antidepressants. I've already shown how his statistics do not represent the data. Garbage in, garbage out.
Kirsch is a psychologist. His specialty is deception, suggestion, hypnosis. I cannot believe he did not consider what he did not say in this paper, along with what he did say, and their respective impacts on the reader.
He has argued that drugs are no better than placebo (even though the evidence is quite clear that they are), and NICE demonstrates that there is modest evidence that psychotherapy is equivalent to drug treatment. Using his tortured logic, but using far better evidence, it would be reasonable to conclude that people should forego psychotherapy, and just take a placebo. He should be out of a job.
Lar
poster:Larry Hoover
thread:895119
URL: http://www.dr-bob.org/babble/20090524/msgs/897569.html