In a study of fluoxetine (Prozac) for adolescents, researchers found that the placebo effect predicted good outcomes, but the actual drug treatment did not. After accounting for “treatment guess” (those who figured out that they were receiving an intervention rather than placebo), the drug was not effective in depression treatment.
In fact, those who received a placebo but thought they received Prozac improved more than those who received the drug and knew it.
“Treatment guesses strongly predicted outcomes and may have led to the exaggeration of drug effectiveness in the absence of actual effects,” the researchers write.
That is, the researchers write that the drug is not actually effective in depression treatment (“the absence of actual effects”) but is perceived to be effective by the researchers because the results were contaminated by the placebo effect.
This finding is similar to another recent finding, which referred to the placebo effect as “subjective beliefs.” In that article, researchers found that in three different studies of neuromodulation for depression, participants’ beliefs that they would improve after using the devices was a significant predictor of outcome, but the actual use of the devices was not, compared to placebo. (In a fourth study, belief was a significant predictor of outcome, but so was the actual treatment.)
The current study was authored by Jon Jureidini, Julie Klau, Natalie Aboustate, and Melissa Raven at the University of Adelaide, Australia, and Joanna Moncrieff at University College London, UK. It was published in the Australian & New Zealand Journal of Psychiatry. Jureidini has been interviewed by Mad in America about his work on the lack of evidence base for psychiatric treatments, including antidepressants for children and adolescents. Moncrieff has also been interviewed by Mad in America, and was the lead author on a 2022 study that put the final nail in the coffin of the chemical imbalance myth of depression.
Fluoxetine is the only FDA-approved antidepressant for use in children and adolescents. Nonetheless, its effectiveness has been questioned, even as it is known to increase the risk of suicide in kids. One recent study found a threefold increase in suicide risk and another study found that the risk may be as high as six times greater. Even in adults, antidepressants may double the risk of suicide.
Placebo-Controlled Studies
In placebo-controlled studies, the group that receives an intervention is compared to a group that does not. But those in that control group are given something (the placebo) that makes them think they have received the intervention. Having this group allows researchers to account for a couple of things. For instance, one is that people may get better naturally, without an intervention, so having a control group that doesn’t receive the intervention allows researchers to compare the intervention to the natural improvement that people experience over time.
But another thing the placebo group controls for is the placebo effect—that people improve when they believe that they have received a treatment, whether they actually did or not. A key feature of this kind of study is blinding, a term for keeping participants, researchers, and/or treatment clinicians in the dark about whether the subject received the treatment or not. This is intended to equalize the two groups: if no one knows whether they received the treatment, then the effect should be consistent across the whole study.
Unfortunately, in this type of study, the blind is often broken. One aspect that can easily break the blind is adverse effects: people know the potential side effects of the drug, so if they experience those effects, they can guess that they are probably in the treatment group, not the placebo group. Those who guess that they are receiving the treatment are likely to have an enhanced placebo effect—to do better because they expect to do better. At the same time, those who guess they received the placebo are likely to do worse because they know they did not receive the actual treatment.
But how much impact those guesses have, and whether the placebo effect is stronger than the actual efficacy of the treatment, is different for every study, treatment, and condition.
The Current Study
Back in 2003, the NIMH-sponsored Treatment for Adolescents with Depression Study (TADS) included 439 adolescents aged 12-17 who met DSM-IV criteria for depression. There were four treatment arms, including fluoxetine (Prozac) only; cognitive-behavioral therapy (CBT) only; fluoxetine and CBT; and placebo. The psychotherapy groups were not able to be blinded. The randomized trial lasted for 12 weeks, and participants were asked which group they believed they were in at both 6 weeks and 12 weeks. Improvement in depression was measured with the Children’s Depression Rating Scale–Revised (CDRS-R).
The TADS study is commonly cited as evidence for Prozac’s effectiveness in depression treatment, because the combined drug and CBT group did slightly better than the placebo group. However, the drug group alone did no better than the placebo group in the TADS analysis of the CDRS-R.
The current analysis was conducted as part of the Restoring Invisible and Abandoned Trials (RIAT) initiative, which allowed the researchers access to the raw data from the TADS study. They obtained information on the fluoxetine group (109 participants) and the placebo group (111 participants), since those were the two blinded groups, in order to directly compare the effects of unblinding.
In all the groups, more than 60% of the participants and raters correctly guessed whether they received the drug or placebo (a perfectly blinded study would result in 50% guessing correctly).
The researchers found that the placebo effect was stronger than the actual treatment itself. Those who guessed that they received the treatment were more likely to improve than those who guessed they received the placebo—even if their guess was incorrect. That is, on average, those who believed they received the drug improved, even if they actually received the placebo. Likewise, those who believed they received the placebo were less likely to improve, even if they actually received the drug.
Those who believed they received the drug, on average, improved by 10 points more on the CDRS-R than those who believed they received the placebo. Those who believed they received the drug improved by 26.98 points, on average. Those who believed they received the placebo improved by 16.65 points, on average.
Amazingly, the group that did the best was those who believed they received the drug, but actually received the placebo. These patients did better than those who received the drug and knew it!
“Adolescents who guessed they were on fluoxetine, but were actually allocated to placebo, demonstrated the largest improvement in CDRS-R,” the researchers write.
Finally, the researchers confirmed the initial finding of TADS: after accounting for the placebo effect (treatment guess), the researchers found that taking Prozac did not improve depression.
The researchers write, “Treatment guess had a substantial and statistically significant effect on outcome (Children’s Depression Rating Scale-Revised change mean difference 9.12, p < 0.001), but actual treatment arm did not (1.53, p = 0.489).”
The researchers conclude that unblinding, which amplifies the placebo effect, may be the reason antidepressants typically beat placebo by a slight margin in clinical trials. They add that future studies need to make sure to assess unblinding in order to provide accurate data on drug efficacy.
“Our analysis suggests that the effects that are demonstrated in placebo-controlled trials of antidepressants may represent amplified placebo effects that are a result of the differential distribution of expectancy effects caused by unblinding. Since the expectancy effects are substantial, even a small degree of unblinding might produce an apparent difference between an active drug and a placebo. For future research, there is a clear need for more stringent study designs that systematically record and analyse treatment guesses and assess blindness, and do so early on and repeatedly,” they write.
Moreover, since clinical practice guidelines are based on evidence from studies like TADS, the researchers argue that guideline authors need to reassess the evidence base for their recommendations. Recommending antidepressants on the basis of studies like TADS is poor science.
Indeed, the evidence base for antidepressants—even for adults—is so poor that a recent article, authored by 30+ prominent figures in the field, recommends against their use in all but “the most severe depression.” The World Health Organization (WHO) guidelines agree: “Antidepressant medications are not needed for mild depression,” according to the WHO. There are plenty of other approaches, with fewer side effects, that are just as effective.
****
Jureidini, J., Moncrieff, J., Klau, J., Aboustate, N., & Raven, M. (2023). Treatment guesses in the Treatment for Adolescents with Depression Study: Accuracy, unblinding and influence on outcomes. Australian & New Zealand Journal of Psychiatry. Published online December 21, 2023. https://doi.org/10.1177/0004867423121862 (Full Text)
Editor’s Note: Part of MITUK’s core mission is to present a scientific critique of the existing paradigm of care. Each week we will be republishing Mad in America’s latest blog on the evidence supporting the need for radical change.
I have not seen such a clear description of some of the difficulties around treatment unblinding before… experiencing side effects being a dead giveaway. Interesting.
It’s a bit scary and I am sure frustrating for researchers to know that the margin of an effect is so easily out shone by placebo and blinding effects.
All the more reason why we need an understanding of the actual problem, what the treatment target is… and how the drug directly affects it.
In the case of the mind… which can be modeled as an objects relations computer… why have I never see a discussion along the lines of:
– perception is the (automated) functionality in which we represent what is, from a stimulus or instance of action
– It yields a prediction and orientates our response towards action that will generate satisfaction in how we relate to that prediction
This string of information is the idea we hold in mind as representative of the scene we are in.
– We are able to apply our top down executive function to model plausible outcomes
– We are able to model the effects that ‘next’ actions, my own and/or those of others
– That will accept the prediction or generate alternative outcomes
– A context in which we can assess the implications and value of particular outcomes
– To find the most appropriate or beneficial way to respond
From this ‘set’ of plausible actions we can signal others and work together to find a way to achieve outcomes that are acceptable to the many. Which will generate recognition.
If we get this right often enough, our thoughts and actions seamlessly generate ‘reward’, which we will positively associate with the thoughts and actions taken.
Where are these drugs… including the placebo… acting in this repetitive ‘process’ through which we relate to the world. Why are we talking chemistry rather tha processing?