If it doesn't work, stop it

BMJ 2004; 328 doi: https://doi.org/10.1136/bmj.328.7446.1015 (Published 22 April 2004) Cite this as: BMJ 2004;328:1015

“Ineffective” treatments might be in a patient's best interests

  1. Peter R Jackson, reader in clinical pharmacology and therapeutics (Peter.R.Jackson{at}shef.ac.uk)
  1. Royal Hallamshire Hospital, Sheffield S10 2JF

    EDITOR—To have a themed issue (28 February) on negative trial results was innovative but risks returning to a dichotomous view of outcomes that should have been abandoned with hypothesis testing. Even means and confidence intervals ignore the wide range of responses likely to be experienced by individual patients.

    Much of this variation probably represents measurement error or intraindividual variability, but some will be due to true heterogeneity in response. Given a symmetrical distribution of outcomes, 50% of patients will benefit from a treatment with a convincingly null effect (mean zero and confidence interval excluding clinically significant harm or benefit). Most individual beneficial responses to an “ineffective” drug will be small, but some could be large enough to be clinically worthwhile even when the confidence interval for the mean apparently excludes significant gains. An equal number of patients will of course suffer detriment. The only difference between effective and ineffective treatments is that the proportion of patients gaining clinically important benefit is greater and the proportion suffering significant harm smaller with effective treatments.

    Every treatment is experimental. When treatments of equal tolerability, safety, and cost are available only a fool would choose the “ineffective” treatment because of the smaller chance of benefit and the greater chance of harm. However, some patients might prefer to try an “ineffective” non-drug treatment to “effective” treatment based on “chemicals.” This would seem rational if any harm from choosing such treatment is reversible, and it illustrates how outcomes usually measured in clinical trials might not be those of primary interest to patients. It does, however, lay on the doctor an even greater responsibility to monitor treatment and ensure any beneficial response persists. In this limited context I disagree with your urging to stop ineffective treatments.


    • Competing interests None declared.

    View Abstract