Intended for healthcare professionals

Education And Debate

Commentary: Placebo run ins have some value

BMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7088.1193 (Published 19 April 1997) Cite this as: BMJ 1997;314:1193
  1. Lawrence E Ramsay, professor
  1. a Section of Clinical Pharmacology and Therapeutics, Department of Medicine and Pharmacology, Royal Hallamshire Hospital, Sheffield S10 2JF

    Commentary

    Professor Senn is right to question placebo run ins, which have tended to become a ritual in clinical trial protocols. I agree with many of his points, but not all of them, and suggest that placebo run ins should not be abandoned entirely.

    Statistics

    Senn contends that the statistical or scientific case for placebo run ins is weak. I agree that they should not be used generally to weed out non-compliers, responders to placebo, or patients with above average variability. Their use in this way limits the extent to which the study results can be generalised to ordinary practice. I agree entirely that active treatments should not be compared back to baseline values. If the true efficacy of an active treatment is to be measured then a parallel group of patients treated by placebo must be included. There is value, however, in having baseline measurements in the untreated or placebo treated state, particularly when the condition under study is defined by a continuous variable–for example, hypertension defined by measurement of blood pressure. This confirms that all the patients entered actually have the condition under study–for example, hypertension of a predetermined degree. It can be determined whether randomisation has yielded treatment groups that are similar. The statistical power of the study is enhanced by comparing changes from baseline between the treatments, as this removes in part variability between patients. Finally, if baseline values do differ between treatment groups to an important extent despite proper randomisation, the situation can be retrieved by statistical analysis using baseline values as a covariate.

    Can active treatment during the run in period serve the same purpose? Sometimes we will have to make do with this, for ethical reasons. However, this introduces an additional source of variation into the comparison of treatments within the study: the variation between patients in the effect of withdrawing the active treatment. Furthermore there is in theory a difference between the ability of a drug to induce a response de novo or to maintain a response induced by active treatment in the run in period.

    Should the run in treatment be single blind? Any blinding in the run in phase is so fragile as not to be blind at all, and I would not argue strongly between no treatment, non-matching treatment, or matching placebo. The main value of placebo treatment is perhaps to familiarise all concerned (patients, investigators, pharmacists, etc) with the procedures that are to be used throughout the study.

    Ethics

    Senn is concerned with the ethics of placebo run ins. He accepts, as I do, that use of placebos in randomised controlled trials may be uncontroversial in some circumstances. I would define these as a negligible risk to the patient; discomfort that is acceptable to a fully informed patient; and agreement by an independent ethics committee that these conditions hold. Senn is particularly troubled by the element of deception in the single blind placebo run in, because the “open protocol, hidden allocation” condition is not met. I disagree with this, with the important proviso that the informed consent must be appropriate. One form of words might be: “During this study there will be one or more periods during which you will have inactive (placebo) treatment. It is important for the success of the study that you are unaware which study periods these are.”

    Placebo run in periods should not be included in protocols as a ritual. Each study should be considered individually, using the following criteria:

    • Does the use of placebo have a negligible risk and acceptable discomfort?

    • Will a placebo run in period enhance the science or interpretation of the study?

    • Is the “deception” covered adequately in the information provided to patients?

    View Abstract