Thursday, January 7, 2010

Antidepressants Ineffective? Don't Believe It! (Part 1)

A recent study in JAMA (the Journal of the American Medical Association) appeared to show that antidepressants do not work well in patients whose depression is mild to moderate, which is the majority of patients. The authors combined the results of six studies (three evaluated Paxil and three evaluated the old tricyclic imipramine) and found that there was little difference between active drug and placebo in patients with less severe depression. However, there was a large benefit of the active drug in patients with very severe depression.

The study made news--see, for example, this piece in the New York Times. But do the results really mean that antidepressants are ineffective? I don't think so. In order to understand the implications of the study, you have to understand how clinical trials are conducted, and how radically they differ from usual care.

First, let's look at what happens if you were to seek care through the usual treatment system. You have become increasingly morose after your 401K tanked in the current recession. For weeks, you have been unable to sleep well, you are wracked with anxiety and panic, you have no energy, nothing is enjoyable, and you have been questioning whether life is worth living. You go to a psychiatrist, who diagnoses you with major depression and anxiety and prescribes you a combination of Zoloft (an antidepressant) and Klonopin (an antianxiety medication.) She tells you that your symptoms are an understandable response to a difficult situation, and that in her experience the Zoloft/Klonopin combo is almost certain to help over the next few weeks. And indeed, you get better.

Suppose, instead, that you sought treatment through a research study that was advertised on the radio. Your ears perked up when you heard that treatment was free, and that you would be paid a small stipend to attend sessions. You go to your first appointment and you are surprised to find that this is not a university, but a private clinical research company. The majority of clinical trials now occur in such for-profit enterprises, because drug companies have found that they can exert more control over the research conducted at such companies than they can at universities.

You meet first with a research assistant, who asks you a few screening questions, then you meet with a psychiatrist. Unbeknownst to you, the psychiatrist will get paid several thousand dollars for each patient he enrolls in the study. The drug company, which designed the study, has given the psychiatrist a long list of inclusion and exclusion criteria. For example, you are ineligible if you have bipolar disorder, suicidal ideation, panic disorder, obsessive compulsive disorder, substance abuse within the last 6 months, or an unstable medical condition. There are so many exclusion criteria that only about 10% of depressed patients in a typical clinic would be eligible for such studies (see here for more background).

The screening psychiatrist will interview you with a goal of getting you into the study, and will typically try to "undiagnose" you of any conditions that might otherwise exclude you. For example, the anxieties that you have had are very similar to panic attacks, but this is a matter of clinical judgment. After asking several questions, your psychiatrist convinces himself that you do not have panic disorder, but rather, you have anxiety that often accompanies depression. You admit to drinking a little more than normal as a way of coping, but after careful questioning, your psychiatrist opines that your drinking pattern is not quite severe enough to be labeled "alcohol abuse."

And so on. After an hour or so, your psychiatrist tells you that you are eligible to enter the study....

(To be continued. Tune in next week to follow the adventures of our patient who learns about the peculiar world of the placebo-controlled clinical trial).


Cheryl Fuller, Ph.D. said...

Isn't it quite likely that the patient in your first example, coming through the usual treatment system, would have felt better in a relatively short time anyway? Situational depression is usually time-limited, is it not? So, what did the job -- the medications or time?

Cheryl Fuller, Ph.D. said...

Ack -- I meant "self-limiting" not "time-limited" in my previous comment.

Daniel Carlat said...

Cheryl, I agree that we can never know what the active ingredient is--the medication or the placebo effect or the passage of time. My point is to show that placebo controlled clinical trials present such a contrived and artificial environment that they are barely generalizable to real life conditions. I'll flesh this out more next week.

Gallucci-Neto said...

That clinical trials are artificial there is no doubt. Also they do not represent real life conditions. It is an unsolved question. So the results cannot be simply "exported" to clinical practice. But the study is well design from a clinical perspective and show us the limitations of ours psychometric instruments (HDRS)to measure depression.

qetzal said...

Your argument here in Part 1 seems somewhat self-contradictory. On the one hand, you state that the long list of inclusion/exclusion criteria means only 10% of depressed patients in a typical clinic would be eligible. But then you argue that the study PI will find ways to enroll people anyway.

If the I/E criteria mean that trial participants aren't really comparable to the patients who get treated in practice, then I agree that the trial result may not be applicable. But if the study PI is actually finding ways to enroll people more generally, by "undiagnosing" them of potential exclusion criteria, then the trial participants should be more comparable to real-world patients.

So I'm not sure where you're going with this argument. Are you suggesting that the trial might have enrolled subjects that weren't depressed, and wouldn't have been treated with these drugs during normal clinical practice? That would certainly invalidate the results of the trial, but it would imply something approaching misconduct on the part of the study PI. Either that, or the criteria for major depression are so vague as to be nearly useless.

Or are you suggesting that practicing clinicians will actually adhere more closely to the prescribing criteria from the trial? That seems quite unlikely to me.

Anonymous said...

Dan having worked on antidepressant drug trials in the pharmaceutical industry for years I can tell you that the findings of this study are absolutely true.

There is a reason drug companies never do studies in mild depression. Because we know we would never see any positive results. The studies for approval are all done in outpatients with moderate depression (i.e. HAM-D >/= 18 or 20 and do not include patients with severe depression (HAM-D ~ > 28 and requiring hospitalization). In Europe a study in this severe inpatient population is needed for approval in some countries and so a single study in this population is often also done but it is not part of the basis for approval in the US.

We know that when we do the studies in moderate depression we are more likely to get a negative study if we enroll more subjects at the lower end of moderate depression and more likely to see a positive response if we enroll more subjects at the high end of moderate depression. You can see this if you look at the mean HAM-D score in positive and negative studies.

The criteria for a positive response for a subject in an antidepressant trial is a >= 50% drop in HAM-D score and a final HAM-D score of 12 or less.

One reason for this lack of being able to demonstrate efficacy in patients with milder depression is that there is always a degree of improvement on placebo. A second reason is when you enroll less sick subjects it is harder to get a response in a subject that qualifies them as a responder.

For example if all your patients have HAM-D scores of 18 - 20 you need their final scores to be 9 or 10 to be considered a responder. But if you enroll subjects with scores of 24 - 26 you only need to reach a HAM-D score of 12 to be considered a responder. If the final score in patients on drug who are responders is always 10 -12 regardless of baseline score and if the placebo response is always a drop of 6 points regardless of baseline score. Then it's clear you are more likely to have a positive study based on the difference in the percentage of responders as defined by the regulatory criteria if you enroll sicker subjects, as well as if determined as the difference in the final average HAM-D scores.


J Hassman, MD said...

I see more and more of late being asked to treat socioeconomic conditions, not psychological ones on their merit alone. And these patients come in DEMANDING meds, not problem solving with the clinician to seek out a comprehensive treatment plan to have a greater likelihood of impacting on their needs and issues.

This post disturbs me, as I read it to somewhat validate the process, even though the author is posting it to challenge the article about less efficacy of meds. Go over to Furious Seasons and read a recent post, this week I think, about the increase in polypharmacy going on in psychiatry.

Sorry, I just hope there ends up being a malpractice standard that has validation for those psychiatrists who just feed this meds only sell out! Come on, 20-30 year old patients with a mood disorder diagnosis on 4 or 5 psychotropics?! Obscene is an understatement. And I am a board certified psychiatrist saying this!

harpy said...

"You have become increasingly morose after your 401K tanked in the current recession. For weeks, you have been unable to sleep well, you are wracked with anxiety and panic, you have no energy, nothing is enjoyable, and you have been questioning whether life is worth living"

This this a valid diagnosis of major depression? What's all that bunk about brain chemicals being out of whack, then? Of course the poor bugger is depressed. How is doping himself up gonna help him in the long run? How long will he have to take these drugs to feel ok? What about therapy? Aren't they supposed to be used in conjunction with therapy? What about exercise and voluteering and getting over your poor, pathetic self?

No wonder they hand this shit out like candy. I'll tune in next week, but I must say I'm woefully unimpressed so far.

Joseph Arpaia, MD said...

I think the accurate diagnosis would be Adjustment Disorder with Depressed Mood, or Adjustment Disorder, with Mixed Anxiety and Depressed Mood. Note that the diagnosis of adjustment disorder is valid even if the symptoms would be considered normal under the circumstances, i.e. it is a disorder because it is interfering with the patient's life, not because the patient is disordered.

I think that the vast majority of patients with anxiety and depression really have an adjustment disorder. And it may take several sessions before the patient trusts me enough to reveal the stress (my boss is trying to force me to have sex with him, for example). The symptoms are an ineffective response to the environmental stresses, which may be acute or chronic.

I also think that the mental health field has ignored the idea that anxiety and depression are responses to the environment in favor of the ridiculous "chemical imbalance" theory. I have my own guesses as to why this is so.

Koch's postulates on the cause of a disease are, in my experience, met far more by environmental factors than by chemical imbalances.

Medications seem to blunt the neuroendocrine responses to stress, which provides some relief, but does not change the global behavior of the system and without such changes the symptoms reassert themselves.

Sarah said...

Are antidepressants often prescribed in the situation described above?

I don't mean to minimize, in any way, the distress of the hypothetical patient. It's just that it had not occurred to me that the use described above might be a common one, and so i'm surprised.

Anonymous said...

The article in question evaluated only paroxetine and imipramine. Many of the included studies in the mega-analysis involved a psychotherapy arm. It is a reasonable assumption that people willing to enter a trial containing potential randomization to psychotherapy alone may have characteristics of their depression different from those of patients looking for medication. Patients often have a pretty good idea of what they need to get better.

Moreover, chronicity of symptoms was not considered in this article (as the authors point out themselves). Antidepressants are clearly effective for dysthymia (long-lasting mild depressive symptoms). Thus, the authors over-reach in their conclusion that antidepressants are ineffective in mild to moderate depression. Brief duration of depression with mild to moderate symptoms may be the most placebo-responsive type of depressive syndrome. More persistent symptoms (regardless of severity) or quite severe symptoms resulting in significant role impairment are probably the most medication-responsive depressive syndromes.

As always, the devil's in the details.

Joseph P. Arpaia, MD said...

Anonymous who posted this morning,

There are two previous meta-analysis on this topic which reached similar conclusions.

One of them is here:

The problem with our field is that most of what we are being asked to treat is very imprecisely defined and there are numerous confounding variables which cannot be controlled for in any study.

For example, no one looks at the fact that about 80% of the patients in a study are able to guess correctly what they are taking. This renders the double-blind design useless and skews the placebo effect toward the group taking the medication. This bias is not even looked at, much less corrected for, and it is only one confounding factor of many.

One cannot do science without controlled experiments. The idea that we can answer questions about depression anxiety or most psychiatric disorder scientifically is a delusion. I think we should acknowledge that instead of trying to cloak ourselves in a mantle of authority which is not cut for us.

Neuroskeptic said...

I've written about this and other related issues previously.

My conclusion being that we have no idea how well antidepressants in the great majority of people who currently take them. Not good.

Ana said...

Are you recruiting patients for clinical trials?
I don't understand why you are so concerned with clinical trials.
You have been prescribing SSRIs for how long?
Can't you tell us what your clinical experience show?
How many patients were helped be Zoloft on a long-term basis?
Have you ever helped a patient withdraw clonazepam? If so what approach did you take and for how long the withdrawal lasted?
Do you listen to your patients and if so do you accept their views or try to show them that they are not right and that it is impossible that the panic attacks started after clonazepam prescription.
Please, Carlat, if you want to make any difference and really help your patients you have to go beyond clinical trials.
You should be concerned with the Phase IV of clinical trials.
It is up to you to make the post-market surveillance and it seems you don't care about it.
I tried to search your other site but I don't have money to pay.

Anonymous said...

After trying several of the top ten antidepressants and never feeling exactly right - and gaining at least 30 pounds per episode - after only 2 weeks on Pristiq I have never felt better and have energy and the desire to live and exercise.

Anonymous said...

I don't understand. if the anti-depressants work for people that re severely depressed why would they not work with the moderately depressed? I think that these do work for mild and moderately depressed people if anti depressants release serotonin to make up for the lack of serotonin in the patient then surely it should help. There is one thing I do not understand some antidepressants have a warning that the patient may become more depressed or suicidal. Is this because of the chemicals released reacting badly or is this because these antidepressants do not work and have to be taken by people who actually want to get better and can do so though belief? From my limited knowledge on clinical trials is that they don't really show how things actually are and how things work in the real world. u15224912