Medications: Are Blind Trials Really Blind?

We've been talking about psychopharmacology and the testing of medications.  In the last post on this subject, we considered the idea of the placebo control trial. But what about this "blind" business?

As you'll recall, the point of a double-blind placebo-controlled trial is that neither the treating clinician nor the patient knows which condition the patient is in. 

Why is this important? 

Patients’ belief that they are being treated with a medication can cause them to expect to get better. As a result, they may notice signs of improvement, even if the drug doesn’t really do anything. So if we give one group the “brand new drug” and the other group nothing, then the treated group will almost always do better.

Instead, we give both groups something that looks like a treatment. In fact, it must look identical (because some research suggests that certain pill colours and sizes are seen as more potent than others). If there is an expectancy effect, then it should appear equally in the groups getting the placebo and the active medication. Any differences in the effect between the groups would have to be due to the active ingredient in the pill.

Remind me why the clinician has to be blind.

We all communicate nonverbally. A physician who hands us a prescription confidently will produce more of an expectancy effect (“This’ll clear up your problem!”) than one who exhibits shame, anxiety, or an apologetic manner (“Sorry I don’t have anything more effective to offer you”). If the clinician knows which medication a patient is receiving, he or she might behave differently with clients receiving the real thing than with those receiving the fake. So it’s vital the clinician not know.

Similarly, it’s important that whoever is doing the evaluation of the patient (whether that is the treating clinician or someone else) not know which group the patient is in. If the evaluator really hopes that the test drug works, people who got the drug may be scored as more improved than those who got the placebo – even if the evaluator is trying hard to be even-handed.

So with all these precautions, patients must not know what drug they’re getting, right?

You’d think so. But no. The problem is that there are no perfect medications that change the target symptom (depression, psoriasis, heart rate) and nothing else. Virtually every medication has side effects.  

If you are in a drug trial and you notice that your hair is dry, you have a metallic taste in your mouth, and your libido is gone, you might very easily guess that the pill you are taking isn’t just a jelly bean. In fact, in antidepressant trials patients are sometimes asked at the end of the study “So, any guesses whether you got the test drug or the placebo?” Antidepressants are notoriously “buggy” with side effects, so patients tend to get the right answer. As a result, the supposedly blind trial isn’t blind at all. 

If patients are getting all kinds of side effects they will strongly suspect they are on the active medication, and they may experience a stronger expectancy effect as a result. “I must be getting the real thing so maybe I really will get better.”

There is a problem even if the drug has no side effects. Imagine a stunningly effective drug that eradicates the problem symptom completely. Patients in the medication condition will notice the vast improvement in their condition and will realize that they must be on the real thing, effectively unblinding the study. But this will be a result of the real effectiveness of the drug. A bigger problem is when the side effects give the story away – potentially making an ineffective medication look better than it really is.

So what?

So even in a well-designed placebo control trial, the study may not actually be blind. As a result, patients may have a greater expectancy effect in the genuine-drug condition than in the placebo condition. The study will show that the drug is more effective than placebo, when in reality the drug condition only had a stronger placebo effect than the sugar pill.

How big a problem is this?

No one knows.

Is there any solution?

Yes. Use a placebo that has side effects. Instead of a sugar pill, we can use an existing medication that is not thought to have an effect on the condition that is being studied. So if we are looking at, say, depression, we could compare our new antidepressant with a drug that produces similar side effects but that has never been shown to be effective against depression. These are called “active placebos.”

Does anyone do this?

It's not done often. In 2004, the Cochrane Collaboration (an organization that conducts reasonably rigourous analyses of the general trend of medical data) published a review of studies in which antidepressants were compared to active placebos (Moncrieff, Wessely, & Hardy, 2004). They found nine such studies - not a particularly impressive number.  

Only two of the nine demonstrated a consistent statistically significant difference between the antidepressant and the active placebo. When the studies were combined, the overall effect size (the difference between drug and active placebo) was 0.39 standard deviations - a relatively weak treatment effect. When one particularly positive trial was not included, the effect size dwindled to 0.17 standard deviations. These differences are smaller than the differences that are usually found when antidepressants are compared to inactive placebo. 

The United Kingdom's National Institute for Health and Clinical Excellence recently chose an effect size of 0.50 as a lower cutoff to represent an effect large enough to be considered clinically (as opposed to statistically) significant. Not everyone agrees with the cutoff, which is somewhat arbitrary, but these findings for antidepressants fall below it. 

Unlike some reviews that find more powerful drug-placebo differences in severely depressed populations, the studies of inpatient populations produced weaker effects than those of outpatient populations.  

This review of a few studies should not be used to conclude that antidepressants are nothing more than active placebo. But it’s not exactly a ringing endorsement, either. The authors concluded that standard drug trials using inactive placebos ("sugar pills") might overestimate drug effects because subjects often figure out which condition they are in and develop higher expectations when they have side effects than when they do not.  

It would be nice to see more studies comparing antidepressant medication with active placebo, to see if we can demonstrate a genuine drug effect that isn't based on patient expectations. Until we do this, some of our former optimism about the antidepressant drug class must be tempered with a bit of skepticism.

Reference

Moncrieff, J, Wessely, S, & Hardy, R Active placebos versus antidepressants for depression. Cochrane Database of Systematic Reviews 2004, Issue 1. Art. No.: CD003012.