Equipoise is not synonymous with uncertaintyBMJ 2001; 323 doi: https://doi.org/10.1136/bmj.323.7312.574/b (Published 08 September 2001) Cite this as: BMJ 2001;323:574
EDITOR—I was surprised by Sackett's contribution in the cluster of letters about clinical equipoise.1 He attacks the word equipoise, on the ground that it is not used as commonly as the alternative word, uncertainty. Uncertainty, unlike equipoise, covers a range of situations, not just clinical trials. It is the meaning behind the word that is important.
Equipoise has been clearly defined within the paradigm of expected utility theory.2 “Patient equipoise” applies when the expected utilities of comparator treatments are equivalent. This provides a clear and precise meaning to the word equipoise, but if someone does not like the word then he or she should substitute another word with the same meaning. Uncertainty cannot be that word. Uncertainty is the opposite of certainty and therefore covers a huge range of possibilities, from equipoise all the way to certainty.
Gifford makes a relevant point in the same cluster of letters, highlighting the difference between the conditions under which a trial may be conducted and those that are necessary for participation of individual patients.1 So Sackett is right when he says that uncertainty is a perfectly appropriate criterion on which to mount a trial. However, Gifford is right in saying that the amount of evidence required for a policy decision (for example, to approve a new treatment) is much greater than that required for individual patients to choose their treatment (or for the doctor to do so if the patient cannot take part in the decision).
Sackett is right when he says that uncertainty is the most widely used term, and it can describe the conditions under which a trial can ethically proceed. However, patient equipoise or some similar term is needed to describe the conditions under which patients should rationally accept randomisation.
When I counselled patients about amniocentesis and chorionic villus sampling I did not just say that I was uncertain about their effects but gave a range of probabilities within which I believed the true effect would lie, and I disclosed my best prior estimate within this range. I said that amniocentesis would cause miscarriage in 1 case in 200 while chorionic villus sampling had between 1.5 and 4 times this risk and that my best prior guess was that it would be twice as risky. As a result, some patients chose amniocentesis, some (mostly people at high genetic risk) chose chorionic villus sampling, and a few were unable to decide—they were in personal equipoise and went into the Medical Research Council's trial comparing the two techniques.