I’d like to thank everyone for their comments, both online and “offline” about my recent “Placebos as Effective as Antidepressants” article. Some very good questions were raised in a comment from Stephan, author of the highly recommended Whole Health Blog that I would like to address in today’s article.
The tricky thing about doing scientific research, as I explained in last week’s article, is that conflicts of interest between doctors, researchers and pharmaceutical companies have become so prevalent that the results of even studies published in prominent, peer-reviewed journals cannot be taken at face value.
One must ask: was the study designed properly? Do the author’s conclusions match their own data? Have the authors reported all of the relevant results? Who funded the study, and what role did they have in choosing the subjects, overseeing the methodology and publishing the results?
When looking at a body of research, one must also consider whether there are unpublished studies on the topic and what the effect of those studies might be. This is particularly true in the case of antidepressants, where it has been estimated that approximately 23% of studies have not been published. Why? Because those studies had even less favorable results than those studies that have been published, and the drug companies who paid for them are under no legal obligation (currently – hopefully this might change in the future) to publish study results.
With that in mind, let’s consider Stephan’s comment and each of the points he brings up in turn:
“I fully agree with you about the “chemical imbalance” thing; it’s incredibly dense. They try to spin it like you were born with depression and there’s nothing you can do about it but take a drug. I have a friend who’s into mood disorder research and I’ve talked to him about that meta-analysis showing no significant effect of ADs.
First of all, he has no dog in the fight because his interest in mood disorders is purely academic. I can vouch for his lack of bias toward antidepressants. Here’s what he told me. Basically, what we call “depression” is actually a collection of related disorders. Antidepressants only work on a subset of them.
There are “responders” and “non-responders” in any group of people who receive antidepressants. For responders, antidepressants can be very effective. When you do a meta-analysis where you’re averaging everyone with “depression” together, the effect of an antidepressant will be small or nonexistent because of the heterogeneity. So this is a problem with saying that antidepressants aren’t effective based on that analysis.
I don’t have a problem believing that antidepressants work for some people. They certainly work in animal models of depression, where there is no placebo effect. I don’t think we should banish them from planet Earth. But I do think the fact that we use them so much points to a bigger problem that we should be addressing by other means.”
I want to thank Stephan again for his comment and for raising these important issues.
Let’s start with the parts that I agree with. Certainly, depression has become so broad a term that some have argued that it is an essentially meaningless clinical designation.
In their book The Loss of Sadness, Horwitz and Wakefield point out that the diagnosis of depression has now come to include transient and completely appropriate responses to life such as sadness after the passing of a loved one, disappointment after the loss of a job or anxiety about financial troubles. They argue, very convincingly, that the DSM IV criteria for depression do not adequately distinguish between what they call “normal sadness” and depression, and the result has been the almost complete medicalization of our emotional response to life. I will be writing an article on this very soon, as I believe it’s a critical perspective to understand in our exploration of depression and antidepressants.
Secondly, I certainly cannot argue with the statement that “antidepressants work for some people”. However, the important questions to ask in relation to that statement are:
- Why do antidepressants work?
- How do antidepressants work?
- Whom do they work for?
Why are these questions important? Because if it turns out that antidepressants do not have any specific drug effects (raising serotonin, for example), and work only because people expect or desire them to work (like placebos), or because of non-specific drug effects (such as sedation or stimulation), then the logic behind prescribing antidepressants at all becomes quite tenuous to say the least – especially when their side effects and risks are taken into consideration.
Now let’s consider each point in turn.
“Basically, what we call “depression” is actually a collection of related disorders. Antidepressants only work on a subset of them. There are “responders” and “non-responders” in any group of people who receive antidepressants. For responders, antidepressants can be very effective. When you do a meta-analysis where you’re averaging everyone with “depression” together, the effect of an antidepressant will be small or nonexistent because of the heterogeneity. So this is a problem with saying that antidepressants aren’t effective based on that analysis.”
What Stephan says about the dangers of meta-analysis are true. However, the same danger applies to overestimating the efficacy of antidepressants. There may be some people who are “responders” who have a response to the drug that is significantly higher than placebo; however, there may also be “nonresponders” whose response to the drug was significantly lower than placebo. There is no indication that antidepressants work on a particular subset of people, and no one has identified who this subset is and why they are helped by antidepressants when others are not.
Kirsch’s analysis that antidepressants do not have a clinically meaningful advantage over placebo (“The Emperor’s New Drugs“) has been almost universally accepted within the scientific community. Kirsch and his colleagues invoked the Freedom of Information Act to obtain access to the FDA database of controlled trials used in the initial approval for the most popular antidepressants. According to researcher David O. Antonuccio in his article “Antidepressants: A Triumph of Marketing Over Science?“, it is difficult to imagine a database that would offer a more fair opportunity to evaluate the efficacy of antidepressants.
The fact that Kirsch found that antidepressants were no more effective than placebo is surprising, in a way, because these results come from studies that were underwritten and designed by the drug companies themselves under conditions most favorable to the active drug condition. In other words, the deck was stacked from the start in favor of the drugs, and they still didn’t come out ahead.
Since a drug must be shown to be superior to placebo in order to be approved, placebo effects are incredibly annoying to drug companies and they do everything they can to minimize the impact of placebo in their studies. The fact that the FDA allows them to use these techniques is, in my mind, blatant corruption. Consider the following methods used in the studies Kirsch analyzed (and most studies, for that matter):
- Placebo washout period: During the first two weeks of the study, everyone is on a placebo. The subjects that respond best to placebo are eliminated from the study. This potentially removes both antidepressant nonresponders (i.e. those that were on an antidepressant before the study starts who get better when they are taken off of it) and placebo resopnders (i.e. those that are not on antidepressants before the study and who respond to placebo). Imagine the converse: an antidepressant washout procedure that eliminates all of the antidepressant responders before a study begins! Such a procedure would surely be considered bias.
- Penetration of the blind: the double-blind in these studies (where neither the patients nor doctors are supposed to know who is taking the drug and who is taking the placebo) is likely to be unintentionally broken because of the pattern of side effects in the active and inactive drug conditions (Greenberg & Fisher, 1997). When efforts are made to ensure the integrity of the blind, drug effects are diminished. For example, a recent review of the Cochrane database of antidepressant studies using “active placebos” (making side effects more difficult to detect) found very small, non-significant differences between drug and placebo, suggesting that trials using inert placebos overestimate drug effects (Moncrieff, Wessely & Hardy, 2001)
- Replacement of non-responders: at least six of the studies Kirsch reviewed (of 38) allowed replacement of nonresponders. This means that during the first two weeks of the study, those that were not responding to the drug were removed from the study!
- Use of sedative medication: Most studies allowed the prescription of a sedative concurrent with the antidepressant. Since other studies have shown that sedatives are as effective in treating depression as antidepressants, how can we possibly know whether the antidepressant effect obtained (if there was one) in these studies was due to the antidepressant drug or the sedative? There are at least 6 points on the 52-point HAM-D scale that the doctors use to determine whether a patient is depressed or not that are related to sleep and favor medications with sedative properties. Since the mean difference between patients taking drugs and patients taking placebos in the studies Kirsch analyzed was only 2 points, it is entirely possible that this small difference is only due to the effects of the sedatives used in the studies and has nothing to do with a true antidepressant drug effect.
- Reliance on clinician ratings: Given that patients tend to report smaller differences than clinicians (Moncrieff, 2001) one certainly wonders how the pattern of results would change using self-report measures like the Beck Depression Inventory.
Frankly, considering the bias against placebo described above, it is simply amazing that placebo still nearly matched the effects of the drug. Imagine what the results might look like if the trials had been performed without these “anti-placebo” measures!
Also, Kirsch et al. pointed out that the overall active drug effects may have been further inflated because mean results were not reported from several studies that found nonsignificant differences between placebo and active drugs.
The advantage to using the FDA database for analysis is considerable. It contains all of the data from initial trials, published or not, and therefore is not subject to the usual publication bias. (As I mentioned earlier in the article, drug companies simply don’t publish negative results. This is known as the “file drawer” phenomenon, since they probably just stuff those studies in some file drawer hoping they will never be found). Antidepressants are significantly more effective than inert placebos in about two-thirds of published trials. However, in the FDA database which includes unpublished trials, Kirsch found that medication outperformed placebo less than half of the time (in 20 of 46 trials).
Although Kirsch’s study is a meta-analyses, he is also looking directly at the results of individual trials. When less than half of individual trials show any advantage for the active drug over placebo, one must really wonder whether these drugs have any specific “antidepressant” effects.
Now on to the next point.
“I don’t have a problem believing that antidepressants work for some people. They certainly work in animal models of depression, where there is no placebo effect.”
Actually, it’s not the case that antidepressants work in animal models. According to Joanna Moncrieff in her book The Myth of the Chemical Cure, it is rarely mentioned that all animal models of depression produce variable results according to where they are conducted. “In other words,” she says, “they are unreliable.” In addition to this, they fail to select specifically for antidepressants and responses are obtained with drugs that are not generally considered to be antidepressants (i.e. amphetamines, opiates, antihistamines, antipsychotics, atropine, pentobarbital, zinc and antibiotics).
Also, the SSRIs (the most popular class of antidepressant drugs) typically fail to have any response in the forced swim test, which is one of the most common antidepressant screening tests. In this test, rats are placed in a tank of water from which they cannot escape. The time until they give up trying to escape is measured, on the assumption that the state of giving up is akin to depression. It is thought that antidepressants should prolong the time to giving up. However, the SSRIs have no such effect.
Finally, other tests for depression also frequently show that non-antidepressant drugs (sedatives, stimulants, antihistamines, etc) yield “false positive” results.
“I don’t think we should banish them from planet Earth. But I do think the fact that we use them so much points to a bigger problem that we should be addressing by other means.”
I’m not so sure we shouldn’t banish them from planet Earth, to tell you the truth.
Longitudinal follow-up studies (which study the effects of antidepresants over the long term – not just the 6-8 week periods the clinical trials look at) show very poor outcomes for people treated for depression both in the hospital and in the community, and the overall prevalence of depression is rising despite increased use of antidepressants (Moncrieff & Kirsch, 2006). Two studies that assessed outcome in depressed patients treated with and without drugs found that people prescribed antidepressants had a slightly worse outcome than those not prescribed them, even after baseline severity had been taken into account (Brugha TS et al, 1992; Ronalds C et al., 1997). No comparable studies exist that show a better outcome in people prescribed antidepressants.
Depression is more common today than before antidepressants were introduced and the outcome has not improved. Epidemiological trends show that the more antidepressants are prescribed, the more prevalent depression is. Sharply rising levels of antidepressant prescribing since the 1990s have been accompanied by increased prevalence of depressive episodes (Patten 2004) and by rising levels of sickness absence for depression (Moncrieff & Pomerleau 2000).
Finally, there is a growing body of research suggesting that antidepressants worsen the chronicity, if not severity, of depression in many patients. Even relatively short-term exposure to antidepressants has been shown to cause chemical and even anatomical changes in the body and brain that could predispose patients to further depressive episodes (Jackson, “Rethinking Psychiatric Drugs“)
Some might argue that antidepressants are important to stave off suicide in very depressed patients. However, there is no evidence that antidepressants reduce the risk of suicide or suicide attempts in comparison with a placebo in clinical trials (Kahn et al. 2000). In fact, rates have actually increased in some age groups and in some countries despite increased antidepressant prescribing (Moncrieff & Kirsch 2006), and when antidepressant trials have been re-analyzed to compensate for erroneous methodologies, the SSRIs have consistently revealed a risk of suicide (completed or attempted) of between two to four times higher than placebo (Jackson, “Rethinking Psychiatric Drugs“)
And we haven’t even talked about side effects yet! Since this article is already very long, I’ll save that for another day. Suffice to say that these are not harmless drugs and the side effects can be severe and potentially fatal.
So I ask you, fair readers, when you add all of this up, should doctors continue to prescribe antidepressants? Let’s see what researchers who have been studying antidepressants and depression for decades have to say:
The SSRIs produce no effects that look likely to be useful in depression. They cause unpleasant agitation in a proportion of patients and, although it is difficult to prove conclusively, an increase in suicidal and violent tendencies may be associated with this effect. Therefore, I can think of no good reason to prescribe them at all. – Joanna Moncrieff, 2008
Moncrieff has actually suggested that the term “antidepressant” is a misnomer, because the drugs have not been demonstrated to have a consistent and specific effect against depression.
There is no doubt that antidepressants have a biochemical impact on the brain, but the valence of that impact is open for considerable debate, and whether it corrects a chemical imbalance is in grave doubt… One day we may look back and marvel at the stroke of marketing genius that led to calling these medications antidepressants in the first place. Kirsch et al. have demonstrated that just because a pill is called an antidepressant, it doesn’t necessarily make it so. – David Antonuccio, 2002
If antidepressant drugs were the only option for treating depression, one might still be able to make an argument for their use in spite of their lack of efficacy and risks. However, it has repeatedly been shown that aerobic exercise, light therapy, Cognitive Behavioral Therapy, St. John’s Wort, bibliotherapy (prayer) and perhaps acupuncture (more studies are needed) are just as effective for treating depression as antidepressants – with few, if any, adverse effects. In fact, in the case of exercise many of the side effects produced are beneficial (e.g. better overall health and wellness).
Considering that antidepressants are likely no more effective than placebos, have not improved (and perhaps worsened) long-term outcomes, may increase the risk of suicidal and violent behavior and have significant many other significant side effects and risks, including potentially permanent changes in the brain which predispose patients to further depression… and considering that there are well-established alternatives that are just as effective, if not more so, in treating depression with almost no adverse effects and significantly fewer costs, I see no compelling reason to continue prescribing antidepressants.
Obviously many other people are posing this question, particularly in the medical community. Kirsch’s research has been so widely accepted that the debate has not centered around his conclusions, but on the implications of those conclusions. Ironically, it has been suggested by more than one commentator that although we now know that antidepressants aren’t effective, we should continue to prescribe them – if only in an attempt to elicit a placebo effect.
Huh? Let me explain. One major reason people respond to placebo in antidepressant trials is that they expect the drug to work. They expect it to work because of all of the promotion they’ve seen, newspaper and magazine articles they’ve read, and personal testimonials they’ve heard. What would happen if it became known that antidepressants are, in fact, not effective and that they could actually make depression worse. Bye bye placebo effect.
So some researchers and doctors have actually suggested that we should go on promoting the delusion that antidepressants are effective so that people who are taking them will continue to believe that they are working, which of course significantly increases the chance that they will work.
If these drugs were not so potentially dangerous and harmful, and if there were not proven alternatives, I could almost go along with this deception – although it does raise some very interesting ethical questions. However, the drugs are potentially dangerous and harmful, and there are proven alternatives, so I cannot agree with this approach.
In closing, I just want to remind anyone who is currently taking an antidepressant and thinking about stopping that it is essential you do so under your doctor’s supervision. You will have to gradually taper off of your medication – do not stop abruptly!
As always, I welcome your comments and questions.