Senn in China
Stephen Senn
Competence Centre for Methodology and Statistics
CRP Santé
Strassen, Luxembourg
I think the placebo gets a bad press with ethicists. Many do not seem to understand that the only purpose of a placebo as a control in a randomised clinical trial is to permit the trial to be run as double-blind. A common error is to assume that the giving of a placebo implies the withholding of a known effective treatment. In fact many placebo controlled trials are ‘add-on’ trials in which all patients get proven (partially) effective treatment. We can refer to such treatment as standard common background therapy. In addition, one group gets an unproven experimental treatment and the other a placebo. Used in this way in a randomised clinical trial, the placebo can be a very useful way to increase the precision of our inferences.
A control group helps eliminate many biases: trend effects affecting the patients, local variations in illness, trend effects in assays and regression to the mean. But such biases could be eliminated by having a group given nothing (apart from the standard common background therapy). Only a placebo, however, can allow patients and physicians to be uncertain whether the experimental treatment is being given or not. And ‘blinding’ or ‘masking’ can play a valuable role in eliminating that bias which is due to either expectation of efficacy or fear of side-effects.
However, there is one use of placebo I consider unethical. In many clinical trials a so-called ‘placebo run-in’ is used. That is to say, there is a period after patients are enrolled in the trial and before they are randomised to one of the treatment groups when all of the patients are given a placebo. The reasons can be to stabilise the patients or to screen out those who are poor compliers before the trial proper begins. Indeed, the FDA encourages this use of placebo and, for example, in a 2008 guideline on developing drugs for Diabetes advises: ‘In addition, placebo run-in periods in phase 3 studies can help screen out noncompliant subjects’.
However, there is a problem: what do you tell the patient? Do you explain that a placebo will be given during a run-in period? This would clearly vitiate the purpose of having the run-in. Do you lie to the patient and seek to make him or her believe that an active treatment is being given? This is a clear violation of the principle of informed consent. Most trialists seem to opt for a third option: they somehow fudge the issue and hope that the patient won’t guess. But suppose the patient says to the physician ‘I understand that in many trials there is a placebo run-in; is that the case here?’ What does the physician reply? Clearly the trial must be carried out in the hope that this question is not put and the fact that the trialist has this hope shows that informed consent is not being practised.
Quite different is the use of placebo in the trial proper. Here the treating physician can explain to the patient that he or she will be randomly allocated either placebo or active treatment and neither patient nor physician will know. This is a strategy of mutually agreed deception rather than covert deceit and does not violate informed consent.
The standard we should observe in running clinical trials is ‘open protocol, hidden allocation’. We should hope that patients understand what is being done rather than fearing that they may find out. Placebo run-ins are in clear violation of informed consent and it is a scandal that they continue to be used and that regulatory authorities continue to recommend them.
Error Statistics Philosophy 
Stephen: Thanks so much for this. I take it that you regard it as unethical because the patient is not told that, at this point, (a) she is only being tested for compliance ability, and (b) everyone is getting a placebo. Were she told these things, it might preclude a reliable assessment of compliance. But since (you’ve said elsewhere) the goal is determining what the drug can do (not what it is likely to do) then why not tell them they’re going to first be in a run-in trial?*
I’m no ethicist, but it is interesting to have this matter turn on whether some patients are getting the drug. Perhaps having identified some compliants, the new group could include some to be tested for compliance some to be tested on the drug (with some assignment protocol). How to evaluate the result would be a statistical issue.
Aside: It might be interesting to compare if the compliant ones do better than the less compliant ones on the placebo.
*Is the information of average level of compliance (shown by the run-in stage) taken account of in assessing the drug, since you’re always going to have that?
Deborah: I am mainly concerned about (b) although I agree that (a) is also a problem. I think you could and should share with the patient the knowledge that there will be a run-in period. One possible alternative is to have all patients on the active treatment in the run-in. You could even have a design where you randomly allocated patients to placebo or active in a run-in period and then re-randomised those who showed satisfactory compliance. Yet another possibility is to have no run-in at all and this has many attractions.
As regards what you would do with information on compliance in the trial proper, the intention to treat philosophy requires you to ignore it. Compliance is treated as being an inherent property of treatment and what you are examining is the decsion to prescribe (with all of the consequences that entails) either one treatment or another.
Dr. Senn, you give two reasons for the run-in: stabilizing the patient and compliance screening. I don’t understand what “stabilizing the patient” is and how it would be vitiated by a known run-in.
I propose the following protocol not to solve the problem but just to get a better idea of the issues involved: suppose that the patients were informed that the study involves a placebo run-in period with a random duration unknown to both the patient and the physician?
I can see how informing the patients that there is compliance screening would run into the Hawthorne effect. To me it seems ethically sound on consequentialist grounds to not explain this reason for the run-in period to the patient.
Corey, the main problem with your proposed protocol is that it would either be rather difficult to implement or it would lead to post-randomization exclusions. At some point, the patients would need to receive their randomly allocated medication and for that to happen at a random time they would have to make very frequent clinic visits, which would be impratical in most studies. Post-randomization exclusions should be avoided, to the extent possible, at all costs as they can completely vitiate the benefits of randomization. Anybody can be excluded pre-randomization for any reason at all without threatening the internal validity of the study, hence the desirable goal of excluding those who might not comply post-randomization (including non-compliers ultimately just reduces the power of the trial)… however, while this is certainly a desirable goal, I agree with Stephen that a placebo run-in is not a reasonable way to achieve it.
Thank you Mark! This clarifies a great deal for me. How does compliance monitoring happen anyhow?
What is the reason that patients have to visit the clinic to receive their medication? It ought to be possible to deliver the medication to the patient, in, say, lots of three days worth of doses or whatever, during the run-in period. Is employing couriers to deliver medication considered unacceptable?
This would be expensive and a colossal pain in the butt for the people running the trial, but it would seem to afford a placebo run-in that doesn’t impose a great deal on the patients.
Typically, you essentially ask the patient “did you take your assigned medicine” (which, being a placebo, would not allow for blood tests)… there are fancier versions, such as pill counts or electronic bottle caps, but it essentially boils down to trusting the patient (or making unverifiable assumptions).
They wouldn’t necessarily have to visit the clinic. But they would have to receive drug and, in order to be able to exclude non-compliers pre-randomization, someone would have to assess compliance on a regular basis. But, the most important point that I was trying to make is that randomization could not occur before the patient was ready to start taking the first dose of the allocated drug. They couldn’t be randomized prior to the placebo run-in to start taking, say, drug X after 6 days of run-in, because if they were excluded during run-in then they would be a post-randomization exclusion. They would have to be first randomly assigned to 6 days of run-in, with ongoing compliance assessment by someone who actually knows when randomization occurs (otherwise, how could they be excluded pre-randomization, or what was the point), then randomized to actual drug and started on actual drug, all without the investigator knowing! I can’t see how it could possibly work.
“But, the most important point that I was trying to make is that randomization could not occur before the patient was ready to start taking the first dose of the allocated drug.”
Yup, that was the major take-home message I got from your first reply to me.
“…all without the investigator knowing!”
Huh. I’d have thought that general approaches to the logistics of double-blinding would have been well-worked out by now. The obvious (to me) solution is for the investigator to pay a company to handle the randomization, distribute the drug to clinics, physicians/patients, and sequester the treatment assignment info until the end of the trial. A private company could easily specialize in separating the people handling different tasks both physically and “informationally” while collecting and holding data centrally. I’m imagining a one-stop shop for handling the logistics of double-blinding for clinical trials. Does anything like this already exist?
Corey: I think you are right to query stabilising the patient. A common belief is that the patients will be very variable on entering the trial and this variability will reflect (amongst other things) different treatments they are being given. If they are all given the same treatment (in this case placebo) this will help reduce variation between them. This is commonly believed but I doubt it is true.
A third reason I should have given is the myth of the baseline. Many trialists don’t really understand that it is the head to head comparison of treatment arms that is the purpose of the trial. They insist on defining response as the difference from baseline. They then think that this baseline should be an ‘off treatment’ baseline. Hence the placebo.
A related matter is that much of the hype surrounding pharmacogenetics and personalised medicine is based on the mistaken belief that who responds and who does not can be established in this way.
Stephen: How is compliance checked? Not coming in for refills? Showing up with pills leftover?
Corey: I’m thinking stabilise might mean getting them used to the routine to be employed in the real trials.
I was not going to get into the business of discussing whether or not compliance checking works (my opinion is that it is an imperfect art) but concede that it might and point out that even if it was extremely useful from the scientific point of view the placebo run-in was unethical because if involved deception. The expert on the topic of compliance is John Urqhart who together with colleague Bernard Vrijens runs a company that specialises in compliance checking. Such checking is more useful, in my opinion, in the trial proper rather than the run-in even though, if you take an intention to treat (ITT) point of view, compliance is irrelevant. (However, I can concede that there is life beyond ITT.) John and Bernard have written many interesting papers but a nice start is this one by John:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1874400/
Returning to the placebo run-in, a reasonable alternative, in my opinion, is to have all patients on the active treatment during run-in and then randomise half to continue and half to switch to placebo.