Dear Allstat,

A couple of weeks ago I asks for peoples opinions on the use of single
blind placebo run-ins.
The responses were so varied (ranging from suggesting additional ways in
which they might be used to stating that they have next to no scientific
merit)  that I have decided to present the anonymised responses rather
than summarising them.
I also received a number of additional responses that I felt were
answering the wider question on the use of placebos in general.  I have
not included these responses and I hope that this does not offend
anyone.

I am also happy to receive further responses if there are any.

Best wishes,

Gordon

Original post:

I would be grateful of peoples opinions on the scientific merit and of
the ethics of single blind placebo run-ins?

I have read the paper by Stephen Senn and the paper by Lawrence Ramsay
both in the BMJ and also the paper by Martin Evans in the Journal of
medical ethics.  There does however, appear to be a paucity of
literature in this area.  Is this simply because people do not feel that

it is an issue?  The most common reason that I see for using a single
blind placebo run-in is to weed out non-compliant participants and yet I

have been unable to find any recent literature upholding this method.

All comments gratefully received and if there is sufficient interest
then I will summarise the responses for the list.

Many thanks,

Gordon Taylor

Responses:

As with many matters ethical the answer, I believe, is "it depends".

If the run-in period to the trial involves patients coming off their
regular
medicine (e.g. taken to control asthma or blood pressure or whatever)
and
they may be caused harm according to medical expert opinion then placebo

run-ins are not a good idea and should be avoided in the study design.

On the other hand, if there is no previous medication involved, or if
doctors agree there's no "serious" harm done to the patient in
temporarily
having them come off treatment, then the argument for weeding out
non-compliers is that much stronger. One would then get, hopefully,
trial
results for which ITT analysis and per protocol analyses turn out closer
to
each other (than otherwise would be the case if plenty of non-compliers
are
randomised). Of course, in the situation where patients are denied their

regular treatment it would be preferable to have their informed consent
for
this, (in other words no longer single-blind) though I realise the very
nature of securing pre-randomisaton consent is a tricky one, possibly
causing patients to change the very behaviour researchers want to assess

(namely compliance).

One could seek input from LRECs on a trial-by-trial basis, but, as a
rule of
thumb, if in doubt about possible harm to individuals during a
single-blind
placebo run-in period, the best advice would be to abandon the idea. I
suspect that is true for the majority of trials.

Response 2:

I think that you've answered this already.  The run-in is not a
"statistical
design" part of the trial, but a method for "selecting" patients - to
ensure
they have the right kind of the disease, sufficient severity and that
they
are compliant.

I hope that we are all agreed that randomisation should occur AFTER the
run-in - as it is part of the sample selection process?  Though somehow
I
doubt there is a consensus and I have seen randomisation done before the

run-in "for expediency" - this is poor practice.

Response 3:

Another reason would be to have a fade-out of concomitant medication
that is
not allowed during the trial.

Response 4:

I can see next to no scientific merit.  Even your suggested
justification is very weak, as results from a clinical trial that
selects
only compliant patients are not going to be immediately applicable.

Ethically they can only be defended if all patients are crystal clear
that they are taking placebo during the run-in phase.  The research
ethics committee on which I sit has found that some companies are
prepared to amend information sheets to this effect.  (Refusal to
amend means not having access to local patients).  This, of course,
also undermines the weak argument of selecting for compliance as
there is no reason to assume that compliance with a known
placebo is related to compliance with a potentially active
medication.

In practice, the most common justification given to me has been the
perception, whether or not correct, that the FDA prefer there to be
placebo run-in to some studies.

Response 5:

This was discussed at a symopsium on causal modelling at Ghent
university in
June. Some of the group there are quite keen on single blind placebo
run-ins in
order to give more power/precision to causal models. You could look for
work by
Els Goetghebeur on this subject. Personally I think single blind placebo
run-ins
are ethically dubious. Hope this helps.