Back in November, at the Philosophy of Science Association meeting in Pittsburgh, I heard a really interesting talk by Jeremy Howick of the Centre for Evidence-Based Medicine at Oxford University about the challenges of double-blind trials in medical research. I’m not going to reconstruct his talk here (since it’s his research, not mine), but I wanted to give him the credit for bringing some tantalizing details to my attention before I share them with you here.
First, he noted that “blinding” might not be as apt a description of what actually happens in medical trials as “masking”. A double-blind trial is one in which you make an effort to hide, from subjects and experimenters, which subject is receiving the treatment being tested and which is receiving the placebo. However, despite this effort, both subjects and researchers frequently get clues that let them work out who is in which group — so rather than being blind to this information, they are in a situation where the “mask” that is supposed to separate them from this knowledge lets in some light, or has slipped.
Just how effectively masked most clinical trials are is, apparently, a very important (and somewhat controversial) subject of empirical research.
Imperfect blinding is not solely a problem of bad experimental design or execution. Indeed, according to Philip’s Paradox, our most effective treatments cannot be tested in successfully double-blinded conditions. Given that success in a double-blinded treatment is what we’re supposed to be looking for to demonstrate that a treatment is effective, this is, well … paradoxical.
Some examples of treatments resistant to double-blind trials are the Heimlich maneuver, external defibrillation, and water for severe dehydration. Their extreme effectiveness compared to whatever placebo you could put in place unmask the treatment group pretty decisively.
So, how crucial is a double-blind trial in establishing the efficacy of a treatment?
One of the other things that came up in the talk was the notion of an active placebo, which is not completely inert but instead mimics at least some of the known side effects of the drug being tested. As you might guess, one of the ways subjects in clinical trials can peek through the masks to figure out whether they’re getting the treatment or the sugar pill is if they notice some effects — even if those effects feel more like nausea or dry-mouth than like tumor shrinkage or serotonin re-uptake inhibition.
Depending on the side effects, though, you might start to wonder whether placebo administration exposes experimental subjects to more harm than doing nothing would.
Indeed, the worries about successfully masking a clinical trial bring up a pretty significant tension: Getting more objective results seems to depend on withholding information from human subjects, yet researchers have ethical duties to human subjects, whether in the treatment group or the placebo group — ethical duties not only to look out for the health of those subjects, but also to respect their autonomy. And, it’s hard to exercise your autonomy well without having information — maybe including some of the information being withheld from you to blind the experiment.
I don’t think this is an impossible tension. Prospective human subjects are told up front that they may be assigned to the placebo group or the treatment group, and that they (and the clinicians interacting with them) won’t know which. If you don’t feel like that’s enough information for you, you have the option not to participate in the study.
Prospective human subjects are also informed up front of known possible side effects of the drug being tested. This gives prospective subjects a guess at the harms they may be risking by participating in the study, should they end up in the treatment group. If they end up in the placebo group, the harm they are risking is no improvement, or worsening, of the condition that the treatment is intended to treat.
Providing information about possible side effects means subjects have clues to work out which experimental group they’re in … which may contaminate the results of the trial. Not providing this information, though, means falling down on the duty to secure informed consent from subjects.
You might be able to get around this by using active placebos that have the same side effects but don’t treat the condition that the drug you’re testing is hoped to treat. But then, the actual risk undertaken by the subjects in the placebo group is the harm from not treating their condition plus the harm from the side effects.
Is the extra harm from the active placebo side effects an acceptable risk? If so, are human subjects properly informed when they may be exposed to this harm?