_-The-Breakthrough-Antipsychotic-That-Could-Change-Everything.jpg?1729528747 "KarXT (Cobenfy)_ The Breakthrough Antipsychotic That Could Change Everything.jpg")
Do Antidepressants Beat Placebo?
In case you haven't noticed, antidepressants are under attack. Articles such as "The Emperor's New Drugs" and popular books like "Prozac Backlash" point to a growing public sentiment that psychiatrists need to shed a little hubris.
Some of these attacks are little more than sensationalist accounts of problems we are well aware of, such as the SSRI withdrawal syndrome. But others cut a bit closer to the bone-- namely, a building critique of the notion that antidepressants are any more effective than placebo.
In this article, TCR reviews the data that has formed the basis for a wave of anti-antidepressant skepticism over the last few years.
But first, a brief lesson on how the Food and Drug Administration (FDA) bestows its stamp of approval on new medications.
Once preclinical animal testing identifies a potential antidepressant, that drug is ready to enter the three sequential "phases" of drug testing. Phase 1 enrolls a small number of healthy volunteers who are given a range of different dosages of the experimental drug, usually in a clinical environment. They are closely monitored for side effects, and blood, urine, and stool are collected for metabolic studies. Phase 2 is a preliminary randomized controlled trial in a small number of depressed patients, mainly to decide whether the drug merits definitive efficacy testing. Finally, Phase 3 trials are the large-scale placebo-controlled trials we are all used to, often enrolling 100 patients or more.
In order for a new drug to win FDA approval, the company must show evidence of at least two positive clinical trials. These are called "pivotal" trials, because they enroll large numbers of patients, and they show a statistically significant drug-placebo difference. Like the Medical College Admission Test, you can keep trying until you get a result you like. Thus, companies will conduct several clinical trials in order to be sure they end up with at least two that will pass FDA muster. Most of these trials are never published, especially if they fail to show a drug-placebo difference. But the FDA requires companies to give them all of their results, no matter how lukewarm. It was only a matter of time until enterprising researchers used the Freedom of Information Act to get their hands on this data.
Which brings us is to the groundbreaking Khan, et al, article (1). Khan and colleagues obtained every bit of data, published or unpublished, on pivotal studies submitted to the FDA between 1987 and 1997 on the following antidepressants: Prozac, Zoloft, Paxil, Effexor, Serzone, Remeron, and Wellbutrin. There were 48 studies in the FDA's file (an average of 7 studies per drug), but only 45 included enough data to analyze.
The researchers were faced with a lot of data to plow through, and they had to do the best they could with the quality of information provided to them. This meant that neither response nor remission rates were included in their analysis! All they were given were the average Hamilton Depression Scale (Ham-D) scores for the entire samples. By subtracting the post-treatment Ham-D scores from the pre-treatment scores, they obtained measures of symptom reduction in drug groups vs. placebo groups.
And what were their results? A total of 4,510 patients were assigned to the investigational drugs. On the average, their Ham-D scores decreased by approximately 40%. Compare this figure with the 2,805 patients who received placebo in these trials; these patients' Ham-Ds decreased by about 30%.
Now, what exactly do these numbers mean? Remember, these are not response rates: The authors are not saying that 40% of drug-treated patients responded vs. 30% of placebo- treated patients. Instead, they are comparing the average decrease (read as "improvement") in the score of the major rating instrument used in clinical trials of depression. Now, what does a 40% decrease in the Ham-D really mean?
It's not all that clear, because the Ham-D comes in two popular "flavors"- the 17-item version and the 21- item version, both of which were used in the FDA data-set. Depending on the version used, the maximum score would be in the 50 to 62 point range, while the minimum score, for the deliriously healthy, would be zero. Most non-depressed individuals score around a 5 to 7 on the Hamilton, because occasionally we all have our bad days. Above 10 is "mild" depression, but in order to get into antidepressant trials, you have to have at least a "moderate" depression, defined as a Ham-D of 18 or greater. The average pre-treatment Ham-D in the FDA data was unfortunately not reported, but it appeared to be about 25.
Now we have some numbers to work with. The 40% symptom reduction among patients on active meds would amount to 40% of 25, or a 10- point decrease in the Ham-D. Using the same math, the 30% improvement in the placebo group corresponds to a 7.5 point decrease. Thus, one can say that antidepressant drug treatment per se led to only about a 2.5 point improvement in Ham-D scores. Another way to look at this is that 7.5/10, or 75%, of the drug-related symptom reduction was actually due to placebo.
These data do not make our drugs look very impressive. And to add insult to injury, two very bright psychologists published their own, more sophisticated analysis of the FDA data, in which they weighted the studies differently according to sample sizes (2). Using this technique, fully 80% of medication response was duplicated by placebo.
But remember the limitations of these trials. Patients with too mild or too severe depression were excluded, and no significant concurrent or previous other psychiatric illnesses were permitted. The average dropout rates were in the 40% range, further limiting the generalizability of the findings. Most of the trials were short, only 6 weeks in length, and we know that the placebo effect tends to peter out over the long-haul.
Nonetheless, these data should give us all pause, and should remind us how important non-medication factors are in helping our patients to heal. Sharpening our placebo-enhancing skills means spending more time with patients and imparting to them our confidence that they will get better.
TCR VERDICT: Long Live the Placebo!
1. Khan A, Warner HA, Brown WA (2000), Symptom reduction and suicide risk in patients treated with placebo in antidepressant clinical trials. Arch Gen Psychiatry 57:311-317.
2. Kirsch I and Moore TJ (2002), The emperor’s new drugs: An analysis of antidepressant medication data submitted to the U.S. Food and Drug Administration. Prevention & Treatment, 5, Article 23. Available at www.journals.apa.org/prevention/volume5/pre0050023a.html.
General PsychiatrySome of these attacks are little more than sensationalist accounts of problems we are well aware of, such as the SSRI withdrawal syndrome. But others cut a bit closer to the bone-- namely, a building critique of the notion that antidepressants are any more effective than placebo.
In this article, TCR reviews the data that has formed the basis for a wave of anti-antidepressant skepticism over the last few years.
But first, a brief lesson on how the Food and Drug Administration (FDA) bestows its stamp of approval on new medications.
New research shows that 80% of medication response may be due to the placebo effect
Once preclinical animal testing identifies a potential antidepressant, that drug is ready to enter the three sequential "phases" of drug testing. Phase 1 enrolls a small number of healthy volunteers who are given a range of different dosages of the experimental drug, usually in a clinical environment. They are closely monitored for side effects, and blood, urine, and stool are collected for metabolic studies. Phase 2 is a preliminary randomized controlled trial in a small number of depressed patients, mainly to decide whether the drug merits definitive efficacy testing. Finally, Phase 3 trials are the large-scale placebo-controlled trials we are all used to, often enrolling 100 patients or more.
In order for a new drug to win FDA approval, the company must show evidence of at least two positive clinical trials. These are called "pivotal" trials, because they enroll large numbers of patients, and they show a statistically significant drug-placebo difference. Like the Medical College Admission Test, you can keep trying until you get a result you like. Thus, companies will conduct several clinical trials in order to be sure they end up with at least two that will pass FDA muster. Most of these trials are never published, especially if they fail to show a drug-placebo difference. But the FDA requires companies to give them all of their results, no matter how lukewarm. It was only a matter of time until enterprising researchers used the Freedom of Information Act to get their hands on this data.
Which brings us is to the groundbreaking Khan, et al, article (1). Khan and colleagues obtained every bit of data, published or unpublished, on pivotal studies submitted to the FDA between 1987 and 1997 on the following antidepressants: Prozac, Zoloft, Paxil, Effexor, Serzone, Remeron, and Wellbutrin. There were 48 studies in the FDA's file (an average of 7 studies per drug), but only 45 included enough data to analyze.
The researchers were faced with a lot of data to plow through, and they had to do the best they could with the quality of information provided to them. This meant that neither response nor remission rates were included in their analysis! All they were given were the average Hamilton Depression Scale (Ham-D) scores for the entire samples. By subtracting the post-treatment Ham-D scores from the pre-treatment scores, they obtained measures of symptom reduction in drug groups vs. placebo groups.
And what were their results? A total of 4,510 patients were assigned to the investigational drugs. On the average, their Ham-D scores decreased by approximately 40%. Compare this figure with the 2,805 patients who received placebo in these trials; these patients' Ham-Ds decreased by about 30%.
Now, what exactly do these numbers mean? Remember, these are not response rates: The authors are not saying that 40% of drug-treated patients responded vs. 30% of placebo- treated patients. Instead, they are comparing the average decrease (read as "improvement") in the score of the major rating instrument used in clinical trials of depression. Now, what does a 40% decrease in the Ham-D really mean?
It's not all that clear, because the Ham-D comes in two popular "flavors"- the 17-item version and the 21- item version, both of which were used in the FDA data-set. Depending on the version used, the maximum score would be in the 50 to 62 point range, while the minimum score, for the deliriously healthy, would be zero. Most non-depressed individuals score around a 5 to 7 on the Hamilton, because occasionally we all have our bad days. Above 10 is "mild" depression, but in order to get into antidepressant trials, you have to have at least a "moderate" depression, defined as a Ham-D of 18 or greater. The average pre-treatment Ham-D in the FDA data was unfortunately not reported, but it appeared to be about 25.
Now we have some numbers to work with. The 40% symptom reduction among patients on active meds would amount to 40% of 25, or a 10- point decrease in the Ham-D. Using the same math, the 30% improvement in the placebo group corresponds to a 7.5 point decrease. Thus, one can say that antidepressant drug treatment per se led to only about a 2.5 point improvement in Ham-D scores. Another way to look at this is that 7.5/10, or 75%, of the drug-related symptom reduction was actually due to placebo.
These data do not make our drugs look very impressive. And to add insult to injury, two very bright psychologists published their own, more sophisticated analysis of the FDA data, in which they weighted the studies differently according to sample sizes (2). Using this technique, fully 80% of medication response was duplicated by placebo.
But remember the limitations of these trials. Patients with too mild or too severe depression were excluded, and no significant concurrent or previous other psychiatric illnesses were permitted. The average dropout rates were in the 40% range, further limiting the generalizability of the findings. Most of the trials were short, only 6 weeks in length, and we know that the placebo effect tends to peter out over the long-haul.
Nonetheless, these data should give us all pause, and should remind us how important non-medication factors are in helping our patients to heal. Sharpening our placebo-enhancing skills means spending more time with patients and imparting to them our confidence that they will get better.
TCR VERDICT: Long Live the Placebo!
1. Khan A, Warner HA, Brown WA (2000), Symptom reduction and suicide risk in patients treated with placebo in antidepressant clinical trials. Arch Gen Psychiatry 57:311-317.
2. Kirsch I and Moore TJ (2002), The emperor’s new drugs: An analysis of antidepressant medication data submitted to the U.S. Food and Drug Administration. Prevention & Treatment, 5, Article 23. Available at www.journals.apa.org/prevention/volume5/pre0050023a.html.
Recommended
