For the last month or so I've been posting about medications, the proposed mechanism of action of SSRI antidepressants, and the monoamine hypothesis.
Running through all of this has been a discussion of the placebo effect, which seems like a fairly straightforward idea, but like many such concepts it quickly morphs into something more complicated the more you look at it.
So: It seems obvious that the placebo effect is the amount of improvement that we see in people who are given a placebo (an inactive treatment). In fact, this seems so obvious that researchers themselves often report results as though this was true.
Well, it isn't.
In fact, the improvement due to expectancy (the true placebo effect) – is only one of several factors contributing to the apparent efficacy of placebos. Let’s look at a few others.
Regression toward the mean
How do you get into a drug trial? You have to meet a number of criteria, one of which is to have a disorder of a minimum level of intensity. So if the trial is on depression, you have to have a depression score of X or more.
Every measure has a degree of error. Imagine that right now your “true” level of depression could be quantified as a score of 20. If we give you a test to find this out, you might get a score of 16, or 21, or 24. You’d get a score of around
20, but seldom would the test peg you perfectly. This is just like your old math tests. Sometimes you’d luck out and look a bit smarter than you really were; sometimes the test would make you look worse than you really were.
Now imagine that we’ll only let people into the study if they get a test score of 21 or higher. People with a “true” score of 10 are unlikely to get a test score wrong enough to be over 21, so they’ll all be denied entry to the study. People with a true score of 35 probably won’t get a test score lower than 21, so they’ll all be admitted.
But think of what happens to people who score close to the cutoff of 21. Some people with a “true” score of 18 will score 22, and they’ll get into the study group. Some people with a “true” score of 23 will score 19, and they’ll be tossed out. All the people with a true score of 21 and who score a bit too high will be included. Those who score a bit too low will be omitted.
In other words, we have a bias: People close to the cutoff who test a bit high will get into the study, and those who test a bit low won’t.
Imagine that neither the drug nor the placebo does anything at all. We test all of our study participants again. Our extremely depressed folks scored high the first time, and they score high again. The second test is a few points higher for some and lower for others due to the random error in the tests. They average out the same.
But think about our group who started out near the cutoff. We have dropped people who randomly tested low from the study, and included those who randomly tested high. When we retest them, about half will score higher than their “true” score, just like they did before; and about half will score too low. But this time we won’t dump the ones who score too low. It will look as though this group of people got a little bit better.
The result: If we use a cutoff score, and if there is any measurement error at all (and there always is), we will always get a little bit of improvement when we retest the subgroup of people we let into the study. We’ll get that improvement in the real drug condition, and in the placebo condition. And we’ll call it a placebo effect, but it isn’t. It’s a statistical anomaly.
Passage of Time
Again, think of how people get into a medication study for a problem like depression. They’re depressed!
What is the natural history of depression? It gets a bit worse and it gets a bit better. Later on, it can get worse again.
Now: When are you going to seek treatment, or volunteer for an experimental drug trial? Probably not when you’re feeling better. You’d only seek help if you were feeling somewhat worse lately.
So if we conduct a drug trial for depression, we’re going to get a lot of volunteers who have a mood problem that spontaneously gets worse and gets better. And most of them will be at the “worse” phase. If we admit them to the study and give them a medication, or a sugar pill, or just have them wait around, what will happen?
A few of them, who felt fairly lousy, will get even worse. But many of them, who volunteered precisely because they felt unusually bad, will feel somewhat better. If we test everyone and average out the scores, we’ll discover that the average depression score is a bit lower than when the study started. It’ll look like being a part of our study was therapeutic, but it wasn’t. It was the simple passage of time, plus an anomaly caused by people seeking help at the worst part of their mood cycle.
The Hawthorne Effect
Just being part of a study can help you feel better. In the late 1920s and early 1930s a series of studies was carried out at an American electrical plant to see whether light levels would influence productivity. It turned out that productivity improved when changes were made, regardless of whether the light was increased or decreased.
Since 1950, when the Hawthorne effect (named for the factory) was first described, it has routinely been noticed that simply being a part of a study creates changes in the people being examined. People in a placebo-controlled drug trial are likely to improve regardless of what treatment they are given, simply because they are in a trial.
Care and Support
When you are in a drug trial you go through a long series of intake measures, you see clinicians regularly, and people monitor your condition. In some ways the care you receive is much more attentive than you would get in the average physician's office.
During depression people often feel isolated and insignificant, and they usually have difficulty getting out of the house. We know that if we can help people become more active, if we can have them interact with people more often, and if we can help them to feel (rightly) that they are deserving of care and support, this often works against the depression.
In effect, just having appointments and talking with a professional (or for that matter any warm human being) can be a genuine and effective treatment for depression - even if the medication taken as a part of that relationship does nothing.
For people who are away from work, I have often found that doing volunteer work is one of the most potent strategies to help people recover from depression. It provides structure during the day. It provides a reason to get out (and once out it is easier to stay out a bit longer and do some errands or buy groceries). It provides human contact. It gives the person a sense that they have something to contribute. And the person is genuinely participating in outside life. Participating in a drug trial does all of these things as well.
So in addition to whatever the capsules contain, being in a study is therapy itself.
* * *
These are just some of the effects that can contribute to improvement in a person who takes a placebo in a drug trial. And, of course, all of the same effects can likewise contribute to those who are given the active drug.
Some of these effects - regression to the mean and the passage of time - would also appear in a group given no treatment at all. Some (the Hawthorne effect; care and support) would be much weaker in a group that only came in for pretest and post-test screening.
So we might wonder whether a group of depressed clients given no treatment at all would improve, and by how much. This has been examined and yes, people given no treatment do tend to improve - but not a great deal. The effect size in studies of antidepressants have averaged just under 0.4 (Kirsch & Sapirstein, 1998), beneath a commonly-used cutoff of 0.5 to describe a moderate effect. Placebo groups average over 1.1, by contrast.
There are two main reasons to care about this issue.
First, if we carelessly combine a variety of influences into the placebo effect, we make placebos look much more effective than they really are. If we want to think carefully about the placebo effect, we need to separate it from the other effects that sometimes inflate it.
Second, all of the effects above, PLUS the placebo effect, also apply to people given active medications in these trials. It is arguably a little oversimplified to say this, but the actual drug effect is usually taken to be the magnitude of response for patients in the active drug condition, minus the response of those in the placebo condition.
Sometimes the difference between drug and placebo is so minimal that a genuine chemical effect of the drug cannot be found.
And some studies suggest that the effects we attribute to the medication may just be more powerful placebo effects, brought on by the presence of medication side effects: Study participants experience side effects and conclude that they are on the real drug, thus developing greater expectation than those in the placebo condition.
Well, fine. But what's the actual difference between people given antidepressant medication and those given placebos? Stay tuned.