Antidepressants No Better Than Placebo Says A New Study, But It's Really More Complicated Than That...
The best ways to offer treatment for Major Depression and depression variants is a constantly evolving area, and of course, a controversial one too. In yesterday's Wall Street Journal health blog, author Jacob Goldstein published a blog entry with an attention-grabbing title: "Do Antidepressants Work Better Than Placebos?". There isn't much to this piece; it's not very long or detailed. It simply refers to a study recently published in PLos Medicine. However, the issue is compelling and the post has already gathered numerous comments.
The PLoS study, titled, "Initial Severity and Antidepressant Benefits: A Meta-Analysis of Data Submitted to the Food and Drug Administration", analyzed a large collection of drug efficacy data submitted to the FDA by pharmaceutical companies as part of the approval process required by the FDA (e.g., before the drugs could be marketed in the United States). The approval studies involved comparisons between a given antidepressant drug candidate and a placebo, in an effort to show that the drug produced a greater antidepressant effect than the placebo did. The results of the study are truly interesting, suggesting that when all of the drug approval results are averaged together, drug and placebo are statistically indistinguishable in terms of their antidepressant effectiveness for mild and moderately depressed people. The drugs only seem to be superior to placebo in reducing depression symptoms when patients are severely depressed. Even then, the authors of the study do not interpret this superior effectiveness as evidence of anything good about the various antidepressant drugs themselves, but instead as evidence that severely depressed people respond less well to placebos than less depressed people do.
Some of the terms used in the PLoS Medicine article may be unfamiliar to readers, so before making my own comments, I think it wise to offer some definitions.
The study being discussed is a variety of research study known as meta-analysis. A meta-analysis is different than a regular research study. In a regular research study, some intervention is offered to some study subjects (called "experimental" subjects), while other study subjects (called "controls") receive a non-intervention, which nevertheless looks like it might have some effectiveness. In drug studies, the non-intervention is typically a placebo; something that looks like a drug and is presented in the same format as a real drug is presented in (e.g., in pill or injection format) but which contains no active ingredients. Study subjects are measured before receiving either the experimental or control interventions and then measured again afterwards. Subject's pretreatment and post-treatment scores on measures that are relevant to the drug's expected result are then compared statistically, so as to take into account the possibility that any observed differences between experimental and control subjects' measurements might have been due to simple chance or luck. A drug is not considered to have been effective in altering subjects' measurements until the statistical analyses show that the experimental condition was not only produced better measurements (e.g., greater antidepressant effects) than the control/placebo condition, but also that the difference between the groups is large enough that it could not have happened by chance alone.
A meta-analysis is a variety of archival research. No new data is collected in a meta-analysis. The data that are analyzed in a meta-analysis are the results from multiple past regular research studies, and not any new data being collected. Essentially what happens is that meta-analytic researchers gather together all the relevant past research studies they can find that speak to a single issue, create averages of the measurements reported in those studies, and then compare the average effects across the multiple studies (for experimental and control conditions) to see if they are more different than might have happened by by chance alone. The criteria for "might have happened by chance alone" is set as 5% of the time by widespread (and arbitrary) scientific convention. If 10% of the time an observed difference might have happened simply by chance random variation, then though the difference appears to be real, it cannot be said to be trustworthy. An observed difference between measurements has to occur by chance 5% of the time or less before it can be said to be real.
What the authors of the PLoS Medicine article have asserted is that in the case of the mild to moderately depressed patients who participated in antidepressant drug trials included in their meta-analysis, there was not any reliable or trustworthy average difference in terms of depression reduction between drug and placebo. In other words, considered as a very large group, mild to moderately depressed subjects got the same antidepressant effect from taking a sugar pill as they did from taking an antidepressant drug. On average, severely depressed subjects within the very large group did benefit more from the drug condition than the placebo condition, but that might not have been so much because the drug effects were great as that severe depression dampens the ability of the placebo effect to operate. That is the authors' conclusion anyway.
The authors' findings do tend to support their conclusion. Still I can't help but walk away from reading the article with a sense that there is an anti-pharma or anti-present-generation-antidepressant-medication bias at work which has perhaps shaped how the authors chose to present their findings. I'm not trying to defend the pharmas or the FDA approval process so much as I think that the story about antidepressant effectiveness is larger and more complicated than the article's authors have chosen to present it as. By not engaging this larger complexity, they make the issue seem more simple than it is and that feeds the unfortunate polarization I've seen in the way people talk about the issue.
There are an awful lot of people out there who feel victimized by psychiatry or by the pharmaceutical industry; who feel angry, frustrated, taken advantage of, disenfranchised, etc. Though the pain they feel is very real and not to be dismissed, some of these people go too far in my opinion and start to demonize the pharmas and psychiatry and the FDA, etc. They see a conspiracy to harm the public where other people see simply medications of varying effectiveness. A study like the one we're talking about here is going to feel like crack cocaine to such people because it appears to confirm what they already "know", which is that "antidepressant drugs don't work". I'm not at all sure that is a legitimate conclusion to take from this work, however.
I sympathize with the upset many people commenting on the Wall Street Journal blog seem to be expressing. My guess is that many of them are struggling with very difficult and chronic life-degrading problems for which no treatment has seemed to help. To the extent that they've concluded that professionals and pharma don't care about them is also reflective of in some cases literally poor care. Such personal pain is understandable and legitimate. However, just because you wish for there to be a bad guy (e.g., an evil pharma industry, evil psychiatrists or incompetent GPs., etc.) to focus anger and frustration upon doesn't mean that there is really a bad guy there.
The problem with the polarized and absolute "antidepressant drugs don't work" reading of these results is that it is too broad of a conclusion to make. More accurately, we might say something like, "The average person with mild and moderate depression is equally well treated by placebo and an averaged cluster of diverse antidepressant drugs". This begs a few questions however, namely:
- that depression itself is very likely a diverse condition having multiple causes,
- that people are differentially vulnerable to depression based on a variety of biological, psychological and social factors, and
- that different people also appear to be differentially sensitive to the antidepressant effects of different drugs.
Statistical studies are always about grouped results (e.g., they are "nomothetical" in nature). Meta-analytic studies are about doubly grouped results. It is vitally important to evaluate drug effects in terms of their averaged group effectiveness, but it is also equally important and vital to remember that groups don't get depressed; individuals do. Individuals are given antidepressant medications and have to live with the down and up sides of those treatments. What works for one person may not work for another precisely because of the differential sensitivities and vulnerabilities and stressor profiles that each individual brings to the table. Ideographic (individual) results are also very important to keep in mind when talking about what works treatment-wise.
The fact that Paxil or Prozac or Effexor or what-have-you-antidepressant (or a cocktail of various antidepressants and other agents) might be helpful to one person but not to another is actually completely consistent with the findings presented in the PLoS article, but you'd never know that to read that article or the comments that some people have made about it. Unfortunately, some people are less interested in finding the truth that is actually out there and more interested in supporting a truth that they already "know".
For the record, I'm a Clinical Psychologist. I'm not a drug guy; I'm a therapist. I'm a guy who wants to sell you on the benefits of an empirically validated psychotherapy for depression rather than push anti-depressant medicines of any variety on you whenever that is an empirically defensible thing to do. Both Cognitive-Behavioral Therapy for Depression and Interpersonal Psychotherapy for depression have been scientifically studied to work as well or better than many forms of antidepressant medication, at least for mild and moderate forms of depression. They don't have any side-effects so to speak except for being expensive and inconvenient in terms of time demands. They have the added benefit of continuing to dampen down the possibility of relapse (always a strong possibility when dealing with depression) after treatment ends when compared with drug treatments.
Even despite my strong bias to promote psychotherapy approaches, in order to remain responsible, I have to recognize that there comes a severity of depression problem where psychotherapy treatment formats must take a back-burner, supportive role to antidepressant medication and other psychiatric/neurological interventions (such as ECT and deep brain stimulation). There is no question that antidepressant medication is valuable and no real question that it works (at least for more severe depression problems, much of the time but not all).
I'm much more personally interested in how people's comments reveal the positions they've staked out (some polarized, some moderate, all reflecting an important personal story and personal emotions and frustrations) with regard to this sensitive issue of how best to treat depression than I am impressed with the results of the actual meta-analysis. What interests you most about this story?