Islon Woolf MD
Antidepressants, publication bias and bad science
Updated: Jun 7, 2020
In my email on vaccines I was somewhat critical of alternative medicine. In this email, I will turn a skeptical eye towards science-based medicine and scientific studies.
Although the following information is almost 10 years old, it is certainly worthy of repeating due to the gravity of the problem and the lack of awareness amongst doctors and patients. It represents an excellent teaching point that identifies flaws in science and sends a message to patients to re-evaluate their need for antidepressants and other medications.
Prior to 2009 most published trials demonstrated that antidepressants helped depression
A psychiatrist working for the FDA was able to find 23 unpublished antidepressant trials from FDA data
After combining this new information with published trials, antidepressants probably don’t work for depression; especially in the case of mild to moderate depression
Patients should re-evaluate their need for antidepressants
Science needs to re-evaluate our understanding of depression and its mechanisms
Other therapies for depression such as exercise, meditation, and cognitive therapy may not work since they were proven to work by non-inferiority to antidepressants
When researchers or companies only publish positive results and put negative results in the circular file we call this publication bias
Publication bias is an immense problem in medicine that makes useless treatments look effective
Be skeptical when you see a positive study result in any field
Be careful extrapolating the benefit of a medicine in patients with severe disease to patients with less severe disease - aka medication creep
Antidepressant clinical trials
In the 1970’s and 1980’s, new drugs were developed for depression that worked by increasing levels of certain neurotransmitters in the brain. The next step in the development of these drugs was to test them on patients. New drugs are usually tested first on the sickest patients. In this case, it was tested on the severely depressed. This is because the severely depressed are the ones with the most to gain and the least to lose. And indeed, large clinical trials did prove that severely depressed patients did benefit. But this was only the tip of the depression iceberg; drug companies in the 1990’s were trying to get FDA approval for a much larger market - mild to moderate depression (probably half the population of the country!). Companies started to churn out randomized placebo controlled trials to get FDA approval for this indication. 48 of 51 published clinical trials (94%) were positive, proving that antidepressants worked even in mild to moderate depression. Below is a graphical representation of the studies of the 12 major anti-depressants; this is how a practicing psychiatrist would have seen the evidence prior to 2009.
The gray boxes above the red line represent the positive studies and the black boxes below the red line represent the negative studies. Looks pretty impressive to me.
You know the rest - half of your friends started taking these drugs. Prozac, Lexapro, Wellbutrin. However, Erick Turner, a psychiatrist working for the FDA, noticed that a lot of the data the drug companies submitted to the FDA were not being published in scientific journals. With the Freedom of Information Act he was able to recover 23 trials and reverse the results of 11 existing trials from FDA data - now only 38 of 74 trials (51%) were positive. It appeared that not by coincidence, drug companies were holding back publishing the negative trials. Below is the representation of the evidence including the unpublished trials.
Putting all 74 trials together in a meta-analysis, antidepressants do not seem to work for depression. The review and unpublished studies were written up in the New England Journal of Medicine in 2009 making big headlines at the time. Moreover, if you separate the data it appears that the mild to moderately depressed particularly do not benefit.
Only publishing positive study results and withholding negative results is know as publication bias. It is not exclusive to drug companies. Everybody wants to see a positive result. The scientists want positive results - it proves their ideas correct and gives them academia prestige. The journal editors want positive results - people want to read positive studies not negative studies. And the patients and doctors want postive results - a positive study means we can help humanity - and who does not want to do that?
This is a huge problem in medicine. Think of each study itself as a piece of data in a very large experiment. If you record only the positive trials and throw out the negative trials it becomes apparent how egregious this form of bias is - because you are only seeing part of the entire picture of the data. Currently, to combat publication bias, scientists have created clinicaltrials.gov. It forces trials to be preregistered to that website with specifications of the methods and measurements of the study. Failure to publish trials leads to fines and journals are discouraged from accepting articles not complying. Hopefully this will help remedy this major flaw in scientific studies.
Implications for the veracity of medical claims
This should alert you to be more skeptical when a treatment claims to be effective. In fact, there are many other ways in which studies can be biased toward showing positive results - publications bias is just one. When confronted with any medical claim, "this treatment cures cancer", "this is the cause of autism", or "this type of exercise is better than this other type in our trial of 26 subjects”, be skeptical. It is human nature is to claim to know things we don’t know (think of blood-letting for 2,000 years, world religions, or politics). In fact, our culture seems to reward people that make claims over people that are skeptical. Think Dr Oz or Deepak Chopra.
However, this in no way should be an indictment of the scientific method and clinical trials. They are not perfect, but they are still superior to other forms of knowledge. Our brains were not evolved to think in a logical manner. We evolved to make quick decisions to solve simple problems for the benefit of survival. “I hear rustling in the leaves, it must be a lion, I am going to run”. Most of the time it was not a lion; but in that type of scenario it is safer to be wrong than accurate. The scientific method, with its careful observation, blinding, replication and peer review attempts to correct these biases and short cuts. Even though there are still residual biases, like publication bias, it still brings us much closer to the truth than anecdotal evidence, evidence from experts, or evidence from ancient wisdom. With all its flaws, clinical trials are the lesser of the evils.
Yet, this paper was evidence of one of sciences best attributes. Science is self-correcting. This paper was highly critical of antidepressants and was published in the most prestigious medical journal in the world; it has now been cited in over 2000 other scientific papers. This is a sign of science working. The only thing a scientist likes better than being right is proving another scientist wrong. Only the strongest ideas can prevail. And, in contrast to more dogmatic forms of knowledge, when a treatment is proven not to work, everyone eventually drops it. And that's probably science’s greatest strength.
Implications for the understanding of depression
It was thought that depression was caused by low levels of neurotransmitters like serotonin, dopamine and norepinephrine in the brain. Antidepressants work by increasing these neurotransmitters. However, if antidepressants do not work, what is going on? We need to re-evaluate the mechanisms of depression. The brain is complex. There are billions of neurons and trillions of switches. Antidepressants work by increasing neurotransmitters throughout the entire brain. And maybe turning on all the switches is not a good idea? Maybe, in the future, we will have a more targeted approach that changes neurotransmitter levels in specific areas of the brain, and in specific neurons? The Harvard Newsletter has an excellent exploration of the biological causes of depression.
Is mild depression a different disease than severe depression? Is mild depression a disease at all? From an evolutionary perspective could depression have a selective advantage. Emotions can be thought of as mental shortcuts that evolved to direct our behavior. Fear helps us escape from risk, anger leads us to defend ourselves against aggressors and bullies, disgust keeps us away from bad things etc. Maybe depression prevents us from pursuing unattainable goals? We always look at negative emotions as a bad thing, but maybe they are defenses that directed behavior? As the famous social biologist E.O. Wilson said,
“Love joins hate, aggression, fear, expansiveness, withdrawal, and so on in blends designed not to promote the happiness of the individual, but to favor the maximum transmission of the controlling genes.”
And the evolutionary medicine specialist Randolph Nesse
"If you have a mutation that makes you have more children that makes you unhappy, it will spread.”
For further exploration of theories how evolution could shape emotions this is an excellent lecture on YouTube. Start at 21:20.
Medication creep is a phenomenon in which a drug, initially studied in patients with a very severe form of a disease, is extrapolated for use in patients with less severe disease. Patients that are very sick are desperate for help and, therefore, side effects of the treatment are tolerated. Less sick patients, on the other hand, get less benefit from treatment. Less sick patients also have a better baseline quality of life and live longer; thus, even small side effects are unacceptable. For a given medicine, the risk/benefit ratio for more severe disease is often more favorable than less severe disease. For instance, there is a new medication X that cures heart disease but causes cancer 20 years after it is initiated. It would be reasonable to use this drug in a patient who has had 5 heart attacks. You would not, however, use it in a 40 year old patient with only a mild elevation of cholesterol (benefits do not outweigh risks).
Consequently, medications should be avoided in mildly ill patients unless proven to be very effective and very safe in the long term. Antidepressants do not seem to be effective and carry significant side effects; the logical conclusion is that they should be avoided in mildly depressed patients.
The medication creep phenomenon is also seen in the treatment of: diabetes, high blood pressure, high cholesterol, osteoporosis, and glaucoma. Treating too soon and too aggressively in the early stages of disease may cause more harm than good. It is also the reason I recommend cautioning perfectly healthy people from consuming vitamins, supplements, and herbal medications - they have the most to lose from long term or subtle side effects.
Implications for patients with mild to moderate depression
Depression has a significant placebo response and a high rate of spontaneous remission. I can’t tell you how many patients of mine have had only a single episode of depression because they were going through a rough patch (a death or divorce of a loved one), were subsequently started on an antidepressant, and have been on that medicine for decades despite no relapse. The attitude years ago was that these medicines were preventive - like "vitamins for the brain" - they just keep you "stable and adjusted". With the above re-analysis of data showing they are not as effective as once thought, and the possibility of significant side effects such as sexual dysfunction and suicide, the risk/benefit ratio for these medicines is not looking very good for the mildly depressed. If I encounter a happy patient on antidepressants I try to slowly ween them off. But ideally it should be a shared decision making process. If my patient wants to continue them and feels better, it's ultimately their choice. (WARNING: If you have been on an antidepressant for years, feel great, and want to consider stoping it, please call me first so we can discuss. Don’t just stop it. It is possible there is a withdrawal syndrome and we may need to consult a psychiatrist.)
Implications for non-pharmacologic treatments of depression
I am a big fan of exercise, meditation, and cognitive behavioral therapy in the treatment and prevention of depression. BUT, most of the clinical trials used to demonstrate their benefit have been non-inferiority trials. For example, exercise was proven to be effective for depression in a famous study at Duke. But how was this study constructed? Depression responds robustly to placebo so you would need a placebo exercise group. But what is placebo exercise? How would you get patients to think they are exercising when they are actually not exercising? Maybe yoga? Unlike medication trials where placebo is easy to achieve, it is hard, if not impossible, to accomplish this with therapies such as meditation or exercise. Instead, proof that exercise works for depression was accomplished by directly comparing exercise with antidepressants. Some patients were randomized to exercise and some were randomized to medication. Exercise was shown to be non-inferior or equivalent. However, we just learned that antidepressants probably do not work; thus it follows that exercise probably does not work. Can you see the cascade of problems if antidepressants truly do not work? If the benchmark with which all other therapies were validated does not work - all other therapies do not work.